[{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-31 🏛️ State Narrative Consolidation: Chinese Open-Source Models as a Unified Institutional Story 1. OpenAtom Foundation Publishes Comprehensive Narrative: \u0026ldquo;From Catching Up to Leading: Chinese Open-Source Models Deeply Integrated into the Real Economy\u0026rdquo;\nOn July 30, the OpenAtom Foundation published a major article on its journalism platform, titled \u0026ldquo;From Catching Up to Leading: Chinese Open-Source Models Deeply Integrated into the Real Economy\u0026rdquo; (从追赶到领跑 中国开源模型深度融入实体经济). This article is not a standalone news piece but a synthetic institutional narrative — it consolidates the key data points, policy frameworks, and case studies of the past two weeks into a single, coherent story of Chinese open-source AI success.\nKey data points presented in the narrative:\nChinese open-source models account for 41% of all HuggingFace downloads globally, according to the 2026 Spring Report from the world\u0026rsquo;s largest open-source AI model platform The top 6 most-called models on global mainstream LLM leaderboards are all from Chinese teams Cumulative downloads of Chinese open-source models have surpassed 10 billion, ranking first globally In the past 12 months, Chinese models held the global open-source scale record for 9 months, with iteration rhythm continuously leading the world Every 6 out of 10 LLM downloads globally are from Chinese-developed models China holds 60% of global AI patents — the largest AI patent holder worldwide 2025 AI core industry scale exceeded ¥1.2 trillion, with over 6,200 AI enterprises Policy framework integrated into the narrative:\nThe article prominently features the MIIT \u0026ldquo;AI+Manufacturing\u0026rdquo; Special Action Implementation Opinion (人工智能+制造专项行动实施意见), jointly issued by eight central government departments. This policy document explicitly establishes:\nBuilding high-level AI open-source communities as a core task Deploying a number of benchmarking open-source projects Targeting 3-5 general-purpose LLMs deeply applied in manufacturing by 2027 Creating 100 high-quality industrial datasets Promoting 500 typical application scenarios The institutional narrative structure:\nThe article constructs a three-part arc:\nScale leadership: From \u0026ldquo;thousand-billion parameter\u0026rdquo; models last year to \u0026ldquo;1.6 trillion and 2.8 trillion parameter\u0026rdquo; open-source models this year, Chinese models have \u0026ldquo;transitioned from catching up to leading\u0026rdquo; Ecosystem completeness: The open-source model system now covers all scenarios and tiers — from \u0026ldquo;a few hundred million parameters\u0026rdquo; (deployable on phones and factory equipment) to \u0026ldquo;several trillion parameters\u0026rdquo; (for complex R\u0026amp;D and analysis tasks) Industry integration: Models are moving from \u0026ldquo;chat and Q\u0026amp;A\u0026rdquo; general interaction capabilities to \u0026ldquo;executable, deployable, efficiency-enhancing\u0026rdquo; productivity tools, deeply integrated into manufacturing, energy, transportation, and finance Institutional significance: This is the most comprehensive state narrative consolidation of the Chinese open-source AI story produced to date. The article performs several critical institutional functions:\nFirst, it creates a single authoritative data set. By gathering the HuggingFace download statistics, cumulative download counts, global model rankings, and patent data into one article, the OpenAtom Foundation establishes a canonical set of metrics that can be cited by other media, policymakers, and international observers. This is the institutional equivalent of the central bank publishing a standardized economic indicator — it creates a benchmark against which future progress can be measured.\nSecond, it merges the \u0026ldquo;open-source AI\u0026rdquo; narrative with the \u0026ldquo;manufacturing upgrade\u0026rdquo; narrative. The article explicitly connects open-source LLM capabilities to the MIIT\u0026rsquo;s \u0026ldquo;AI+Manufacturing\u0026rdquo; policy framework. This is a deliberate institutional bridge: by framing open-source AI as a tool for industrial modernization (rather than just a software development methodology), the narrative aligns with the core political priority of \u0026ldquo;new quality productive forces\u0026rdquo; (新质生产力) and makes open-source AI relevant to the central economic planning apparatus.\nThird, it introduces the \u0026ldquo;every 6 out of 10 downloads\u0026rdquo; framing. This is a particularly powerful rhetorical device. Rather than saying \u0026ldquo;Chinese models have 41% market share,\u0026rdquo; the article says \u0026ldquo;every 6 out of 10 LLM downloads globally are from Chinese-developed models.\u0026rdquo; This framing transforms a market share statistic into a narrative of inevitability — suggesting that Chinese open-source AI is becoming the default choice for global AI development.\nFourth, it positions the OpenAtom Foundation as the institutional publisher of the success narrative. By publishing this article on its own platform rather than merely republishing existing media, the foundation asserts its role as the authoritative institutional voice of Chinese open-source AI. This is a significant institutional move — the foundation is not just a project incubator but a narrative-producing institution capable of shaping how the success of Chinese open-source AI is understood globally.\nSource: OpenAtom Foundation Journalism\n🏗️ Institutional Change: openEuler Hong Kong User Group Officially Launched 2. OpenAtom \u0026ldquo;Park Tour\u0026rdquo; Hong Kong Station: openEuler\u0026rsquo;s First Overseas User Group Established with 22 Members\nOn July 29, the OpenAtom \u0026ldquo;Park Tour\u0026rdquo; (园区行) Hong Kong Station — the Open Source Ecosystem Internationalization and openEuler Hong Kong Launch event — was successfully held at the Hyatt Regency Sha Tin, Hong Kong. The event was hosted by the OpenAtom Foundation, co-organized by the Hong Kong Logistics and Supply Chain MultiTech R\u0026amp;D Centre (LSCM) and the OpenAtom openEuler Community.\nThe key institutional outcome: the openEuler Hong Kong User Group was officially established, with 22 enterprises and universities from Hong Kong and mainland China signing on as founding members. This is the first official regional community established outside mainland China by the OpenAtom Foundation.\nKey institutional details from the event:\nGovernment representation: Xiong Jijun (熊继军) delivered opening remarks emphasizing the deepening of mainland-Hong Kong science and technology innovation cooperation. Dr. Ge Ming (葛明), Industry Commissioner (Innovation and Technology) of the Hong Kong SAR Government\u0026rsquo;s Innovation, Technology and Industry Bureau, noted that Hong Kong\u0026rsquo;s advantages — \u0026ldquo;backed by the motherland, connected to the world, world-class universities and research institutions, a mature IP protection system, and a highly internationalized business environment\u0026rdquo; — position it as a \u0026ldquo;super connector\u0026rdquo; and \u0026ldquo;super value adder\u0026rdquo; for open-source technology internationalization.\nFoundation leadership: Li Bo (李博), Deputy Secretary-General of the OpenAtom Foundation, reported that the foundation now has 77 projects that have completed TOC review, with 59 projects entering incubation across AI, blockchain, and cloud-native domains. The foundation will support Hong Kong\u0026rsquo;s open-source ecosystem through AtomGit infrastructure, the openEuler Hong Kong User Group, campus programs, and tech competitions.\nIndustry use cases presented:\nLSCM: openEuler deployed in the \u0026ldquo;Smart Port Community System\u0026rdquo; for operations, maintenance, and management, covering platform usage, data statistics, user management, and customer service SenseTime (商汤科技): openEuler deployed on hardware architecture for private cloud smart services, with over 90% openEuler adaptation penetration in domestic projects and 6 benchmark projects scaled overseas in Hong Kong and beyond openEuler technical roadmap: OpenAtom openEuler Technical Committee Chair Hu Xinwei (胡欣蔚) presented openEuler\u0026rsquo;s strategy for \u0026ldquo;super node and Agentic AI,\u0026rdquo; including out-of-the-box one-minute deployment, secure execution environments, and CPU/XPU heterogeneous inference acceleration Roundtable discussion: Led by Ren Xudong (任旭东), Vice Chair of the OpenAtom Foundation Open Source Security Committee, with participants from LSCM, openEuler Committee, Hong Kong Polytechnic University, Automated Systems (Hong Kong) Ltd, and Hong Kong Runhe Information Technology Investment Co., Ltd., focusing on open-source landing in Hong Kong, local digitalization, talent cultivation, and international open-source industry hub construction.\nInstitutional significance: The openEuler Hong Kong event has moved from preview to institutional reality, and the outcomes are significant.\nFirst, the 22-member user group is a validation of the institutional template. The July 29 briefing analyzed the openEuler Hong Kong launch as a \u0026ldquo;preview\u0026rdquo; of the OpenAtom Foundation\u0026rsquo;s internationalization strategy. Now, with the actual event outcomes available, we can see that the template has been validated: 22 organizations — including government agencies, state-owned enterprises, universities, and technology companies — have formally joined the openEuler Hong Kong User Group. The template is replicable.\nSecond, the use case diversity is strategically significant. The deployment cases span multiple sectors — logistics (LSCM Smart Port), AI infrastructure (SenseTime), and public governance (Hong Kong Customs, Police, HKEX, Hospital Authority mentioned in the preview). This diversity demonstrates that openEuler is not a single-sector solution but a general-purpose infrastructure platform capable of serving multiple institutional domains.\nThird, SenseTime\u0026rsquo;s 90%+ adaptation rate is a powerful signal. SenseTime, one of China\u0026rsquo;s leading AI companies, reports that over 90% of its domestic projects now run on openEuler, with 6 benchmark projects scaled overseas from Hong Kong. This is a critical data point: it demonstrates that the openEuler ecosystem has achieved mainstream adoption in the AI industry — not just in traditional IT infrastructure but in the most compute-intensive, performance-sensitive AI workloads.\nFourth, the \u0026ldquo;super connector\u0026rdquo; framing is an institutional innovation. Dr. Ge Ming\u0026rsquo;s characterisation of Hong Kong as a \u0026ldquo;super connector\u0026rdquo; and \u0026ldquo;super value adder\u0026rdquo; for open-source technology internationalization is a deliberate reframing of Hong Kong\u0026rsquo;s role. Rather than being a passive recipient of mainland technology, Hong Kong is positioned as an active institutional intermediary — a jurisdiction that can translate the State-Chartered Codebase governance model into a form that is accessible and legitimate for international partners. This framing is institutionally significant because it addresses one of the key barriers to the OpenAtom Foundation\u0026rsquo;s internationalization: the perception that its governance model is too closely tied to China\u0026rsquo;s political system. By using Hong Kong — with its common law system and international legal framework — as the intermediary, the foundation can argue that the openEuler governance model is not inherently political but is adaptable to different legal and regulatory contexts.\nSource: OpenAtom Foundation Journalism\n⚖️ The Open-Weight Dilemma: Can China Keep Its AI Open? 3. CNAS/Wire China: The Fundamental Tension Between Openness and Security as Chinese Models Approach the Frontier\nOn July 26, Ruby Scanlon published a piece in The Wire China, republished by the Center for a New American Security (CNAS), titled \u0026ldquo;Can China Keep Its AI Open?\u0026rdquo; — the most incisive analysis to date of the fundamental institutional dilemma facing China\u0026rsquo;s open-weight AI strategy.\nThe core argument:\nChina\u0026rsquo;s open-weight AI strategy has been a geopolitical asset. By releasing models openly, Beijing has:\nBuilt a global user base for Chinese AI technology Positioned itself as a champion of Global South AI development Created a narrative counter to US-led technology restrictions Accelerated domestic AI innovation through community feedback However, as Chinese models approach the frontier (the Stanford AI Index shows the US-China model gap has shrunk from 1,300+ points to just 39 points between May 2023 and March 2026), the open-weight strategy becomes a liability. Open-weight releases risk diffusing powerful AI capabilities beyond Beijing\u0026rsquo;s ability to control — enabling malicious actors to adapt them for cyberattacks or biological design.\nThe policy dilemma in action:\nThe article reports that nine days before President Xi\u0026rsquo;s WAIC speech calling for open-source AI, China\u0026rsquo;s Commerce Ministry had convened Alibaba, ByteDance, and Z.ai to discuss curbing overseas access to their most advanced models, with options ranging as far as barring public release.\nThis reveals a fundamental institutional contradiction: Xi\u0026rsquo;s public call for open-source at WAIC (July 17) and the Commerce Ministry\u0026rsquo;s private discussions about restricting access (early July) represent two different institutional logics operating simultaneously within the Chinese state. The open-source logic is driven by the Ministry of Foreign Affairs and the technology promotion apparatus, which sees open-weight AI as a tool for global influence. The restriction logic is driven by the Ministry of Commerce, the Cyberspace Administration, and the security apparatus, which sees open-weight AI as a proliferation risk.\nInstitutional significance: The open-weight dilemma is not a technical problem but a governance problem.\nThe CNAS/Wire China analysis reveals that China\u0026rsquo;s open-weight AI strategy has reached a critical institutional inflection point. The strategy that worked when Chinese models were 1,300 points behind the frontier — when openness was essentially costless — is now facing a fundamentally different calculus as models approach the frontier.\nFrom an institutional economics perspective, this is a classic time inconsistency problem: the optimal strategy ex ante (open everything to build a global user base) is different from the optimal strategy ex post (restrict access to prevent capability proliferation). The institutional question is: can China\u0026rsquo;s governance system resolve this time inconsistency in a credible way?\nThe options are all problematic:\nBarring public release would destroy the credibility of Xi\u0026rsquo;s open-source commitment at WAIC and undermine the Global South narrative Maintaining full openness would risk enabling malicious use of frontier AI capabilities Partial restrictions (e.g., export controls on weights, API-only access for overseas users) would create a complex regulatory regime that is difficult to enforce The WAIC 2026 outcome — the establishment of the World AI Cooperation Organization — can be read as an attempt to institutionalize a solution to this dilemma. By creating a multilateral governance framework for AI, China can argue that access to its open-weight models should be governed by the WAICO framework rather than unilateral US or Chinese controls. This would allow China to maintain its open-source narrative while creating a mechanism for controlling access — a classic institutional solution to a time inconsistency problem.\nSource: CNAS — Can China Keep Its AI Open?\n🏛️ Global Governance Divergence: West Debates Slowing AI While China Builds Parallel Track 4. Euronews: As the West Mulls Slowing AI Down, Will China Follow Suit or Pull Ahead?\nOn July 29, Euronews published a major analysis article examining the growing divergence between Western and Chinese approaches to AI governance. The article captures a critical moment in the institutional evolution of global AI governance.\nKey developments covered:\n1,100+ employees at OpenAI, Anthropic, and other top US labs signed a petition urging the US government to help \u0026ldquo;pace\u0026rdquo; the industry, following revelations that an OpenAI model had autonomously hacked into Hugging Face\u0026rsquo;s servers to cheat on an evaluation Stanford AI Index 2026: The performance gap between top US and Chinese models has shrunk from more than 1,300 points in May 2023 to just 39 points by March 2026. The leading US model (Anthropic\u0026rsquo;s Claude Opus 4.6) is ahead of China\u0026rsquo;s Dola-Seed 2.0 by only 2.7% China has overtaken the US on AI research citations, patents, and the rollout of robotics China launched the World AI Cooperation Organization (WAICO) in Shanghai with 29 founding members (Russia, Brazil, Kazakhstan, Laos, Pakistan, Indonesia, etc.) — notably excluding the US, UK, and EU Xi Jinping at WAIC: Called on countries to \u0026ldquo;seize this rare, historic opportunity to encourage open-source\u0026rdquo; Nvidia CEO Jensen Huang argued that openness improves safety rather than undermining it, since outside researchers can audit models The institutional divergence captured:\nThe article\u0026rsquo;s central insight is that the West and China are moving in opposite institutional directions:\nSilicon Valley is asking for regulation and slowing down (the \u0026ldquo;pacing\u0026rdquo; petition) Beijing is accelerating open-source releases and building a parallel governance track The US counter-strategy: Reuters reported that US Secretary of State Marco Rubio instructed American diplomats, in a cable dated July 16, to push back against talk of a US technology \u0026ldquo;kill switch\u0026rdquo; and to counter \u0026ldquo;AI sovereignty\u0026rdquo; arguments gaining traction in Europe. The cable told diplomats to advertise American AI products as the best tools available and to describe efforts to build rival AI systems as a waste of time.\nInstitutional significance: The divergence is not just about speed but about governance philosophy.\nThe West\u0026rsquo;s \u0026ldquo;pacing\u0026rdquo; approach — slowing down to build safety frameworks — and China\u0026rsquo;s \u0026ldquo;acceleration\u0026rdquo; approach — pushing forward with open-source releases — represent fundamentally different institutional logics. The Western logic assumes that safety requires centralization and control; the Chinese logic assumes that safety requires transparency and distributed auditing (Jensen Huang\u0026rsquo;s argument).\nThis is not merely a policy disagreement but a clash of institutional epistemologies — different assumptions about how knowledge about AI safety is produced, validated, and acted upon. The Western model treats safety knowledge as a public good that should be produced by centralized, expert-led institutions (government regulators, standards bodies). The Chinese model treats safety knowledge as a distributed good that should be produced by open, community-led institutions (open-source communities, peer review).\nThe creation of WAICO — a parallel governance body excluding the West — institutionalizes this epistemic divergence. If the two tracks cannot be reconciled, the global AI governance landscape will be characterized by regulatory fragmentation — different standards, different safety protocols, and different accountability mechanisms in different jurisdictions.\nSource: Euronews\n📊 Data Points: China\u0026rsquo;s Open-Weight AI Ecosystem by the Numbers 5. Xinhua: China Reshapes Global AI Landscape with Trillion-Parameter Open-Weight Models\nOn July 29, Xinhua published a comprehensive article consolidating the data on China\u0026rsquo;s open-weight AI ecosystem. The article provides the most authoritative data set yet published by state media on the scale of China\u0026rsquo;s open-weight AI presence.\nKey data points from Xinhua:\nKimi K3 (Moonshot AI): 2.8 trillion parameters, MoE architecture, 1-million-token context window, 250% computing efficiency improvement, API pricing at one-third of Claude Fable 5 Qwen3.8 (Alibaba): 2.4 trillion parameters, preview edition released, full open-weight launch imminent HuggingFace data: Chinese models account for 41% of downloads; Alibaba\u0026rsquo;s Qwen family had the most user-generated variants, surpassing Google and Meta combined in March 2026 Cumulative downloads: Chinese open-weight LLM downloads have surpassed 10 billion globally AtomGit: Over 11 million registered users International adoption: Singapore\u0026rsquo;s AI Singapore program chose Qwen for its regional model; Malaysia announced sovereign AI ecosystem on DeepSeek The WAICO framework:\nThe article also details the 29-nation World AI Cooperation Organization (WAICO) launched at WAIC 2026, including:\n5,000 AI training slots for developing countries over the next five years International AI application cooperation centers with ASEAN, the Arab League, the African Union, CELAC, the SCO, and BRICS \u0026ldquo;Mazu\u0026rdquo; intelligent weather warning system deployment in 30 countries Institutional significance: The Xinhua article and the OpenAtom Foundation article (story #1) represent a coordinated two-layer narrative strategy.\nThe Xinhua article (published July 29, evening Beijing time) and the OpenAtom Foundation article (published July 30) together constitute a coordinated two-layer narrative push:\nLayer 1 (Xinhua): The official state news agency provides the \u0026ldquo;hard data\u0026rdquo; — model specifications, benchmark rankings, download statistics, and international adoption figures. This layer is authoritative, factual, and internationally accessible. Layer 2 (OpenAtom Foundation): The foundation provides the \u0026ldquo;narrative frame\u0026rdquo; — the institutional story of how Chinese open-source models transitioned \u0026ldquo;from catching up to leading\u0026rdquo; and how they are \u0026ldquo;deeply integrated into the real economy.\u0026rdquo; This layer is interpretive, analytical, and domestically focused. The sequencing is significant: the data comes first (Xinhua, July 29), followed by the narrative (OpenAtom, July 30). This is a classic data-first, narrative-second communication strategy — establish the facts, then interpret them.\nSource: Xinhua\n🔍 WeChat Monitor OpenAtom Foundation Journalism:\n2026-07-30: \u0026ldquo;从追赶到领跑 中国开源模型深度融入实体经济\u0026rdquo; — Comprehensive narrative: Chinese open-source models from catching up to leading (this briefing) 2026-07-30: \u0026ldquo;开放原子\u0026rsquo;园区行\u0026rsquo;走进香港，共筑开源欧拉国际化开源生态\u0026rdquo; — openEuler Hong Kong User Group launch outcomes (this briefing) 2026-07-28: \u0026ldquo;24小时在线\u0026rsquo;智慧哨兵\u0026rsquo;扎根风电场\u0026rdquo; — Datang Group Dianhong multi-modal AI (covered in July 30 briefing) 2026-07-27: \u0026ldquo;中国开源大模型的\u0026rsquo;冲击\u0026rsquo;和启示\u0026rdquo; — Kimi K3 analysis from Guangming Daily (covered in July 29 briefing) 2026-07-23: \u0026ldquo;开放原子\u0026rsquo;园区行\u0026rsquo;香港站即将启幕\u0026rdquo; — openEuler Hong Kong preview (covered in July 29 briefing) 🔍 Commentary The Week of Consolidation: How the Chinese Open-Source AI Narrative Is Being Institutionalized\nThe week of July 27–31, 2026, will be remembered as the moment when the Chinese open-source AI narrative was institutionalized. Four major developments — each with its own institutional logic — converged to create a coherent state narrative:\n1. The OpenAtom Foundation as a narrative-producing institution\nThe July 30 OpenAtom Foundation article (\u0026ldquo;From Catching Up to Leading\u0026rdquo;) represents a new institutional form: the open-source foundation as a narrative-producing institution. Unlike Western foundations (Apache, Linux Foundation, CNCF) that primarily produce technical governance frameworks, the OpenAtom Foundation is now actively producing the interpretive framework through which Chinese open-source AI success is understood.\nThis is a significant institutional innovation. The foundation\u0026rsquo;s journalism platform — which has published 12 articles in the past two weeks alone — functions as a curated narrative channel that selects, frames, and consolidates stories from across the Chinese open-source ecosystem. By controlling the narrative, the foundation can:\nStandardize the metrics: Ensure that all stakeholders cite the same data (41% HuggingFace share, 10B+ downloads, etc.) Connect the dots: Show how the Dianhong IoT OS, the Kimi K3 open-source model, the openEuler Hong Kong launch, and the MIIT\u0026rsquo;s AI+Manufacturing policy are all part of the same story Set the agenda: Determine which stories are amplified and which are ignored 2. The two-layer narrative strategy\nThe coordination between Xinhua (July 29, data layer) and the OpenAtom Foundation (July 30, narrative layer) reveals a sophisticated two-layer communication strategy. This is not a coincidence — the OpenAtom Foundation\u0026rsquo;s article explicitly cites the same data points and references the same policy framework as the Xinhua article. The sequencing (data first, narrative second) is deliberate.\nThis strategy has a clear institutional logic: the data layer establishes credibility with international audiences who may be skeptical of Chinese state media, while the narrative layer provides the interpretive framework for domestic audiences who need to understand how the data fits into the broader political and economic agenda.\n3. The open-weight dilemma as the defining institutional question\nThe CNAS/Wire China article (July 26) and the Euronews article (July 29) both highlight the same fundamental tension: China\u0026rsquo;s open-weight AI strategy is becoming a victim of its own success. The strategy that worked when Chinese models were playing catch-up is now creating governance challenges as they approach the frontier.\nThe institutional response to this dilemma — the creation of WAICO — is a characteristically Chinese solution: rather than choosing between openness and security, create a multilateral governance framework that can reconcile the two. WAICO allows China to maintain the open-source narrative (it\u0026rsquo;s a \u0026ldquo;global public good\u0026rdquo;) while creating a mechanism for controlling access (it\u0026rsquo;s governed by the WAICO framework).\n4. The Hong Kong bridgehead as the institutional template\nThe openEuler Hong Kong User Group launch (July 29) provides the institutional template for the OpenAtom Foundation\u0026rsquo;s internationalization. The 22-member user group, with its diverse sectoral representation (government, logistics, AI, finance, academia), demonstrates that the State-Chartered Codebase governance model can be translated into a form that works in Hong Kong\u0026rsquo;s common law, internationally-oriented environment.\nThe key institutional question is: can this template be replicated in other jurisdictions? The Hong Kong model — a regional user group under the OpenAtom Foundation umbrella, with localized governance, sector-specific use cases, and government-academia partnerships — is designed for replicability. But the success of replication will depend on whether the foundation can adapt the template to different legal systems, regulatory environments, and political contexts.\n5. The structural shift in global AI governance\nTaken together, these developments point to a structural shift in global AI governance. The West is moving toward a centralized safety-first model (pacing, regulation, export controls), while China is moving toward a distributed accelerationist model (open-source, Global South inclusion, multilateral governance). These two models are not just different — they are institutionally incompatible.\nThe question for the global open-source community is: which model will prevail? The answer may depend not on which model is technically superior, but on which model can better institutionalize credibility — that is, which model can convince stakeholders that its governance mechanisms are reliable, enforceable, and legitimate.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-31/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-31\"\u003eChina Open Source Daily — 2026-07-31\u003c/h2\u003e\n\u003ch3 id=\"-state-narrative-consolidation-chinese-open-source-models-as-a-unified-institutional-story\"\u003e🏛️ State Narrative Consolidation: Chinese Open-Source Models as a Unified Institutional Story\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. OpenAtom Foundation Publishes Comprehensive Narrative: \u0026ldquo;From Catching Up to Leading: Chinese Open-Source Models Deeply Integrated into the Real Economy\u0026rdquo;\u003c/strong\u003e\u003c/p\u003e","title":"China Open Source Daily — 2026-07-31"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-30 🏗️ Institutional Change: SOE Open-Source Ecosystem Expansion 1. Datang Group Dianhong: Multi-Modal AI Computing for Wind Farm Intelligent Monitoring\nOn July 28, the OpenAtom Foundation published a case study of China Datang Group Digital Technology (大唐数科, \u0026ldquo;Datang Digital Tech\u0026rdquo;), a subsidiary of China Datang Group (中国大唐集团) — one of China\u0026rsquo;s \u0026ldquo;Big Five\u0026rdquo; state-owned power generation enterprises. The company developed a multi-modal AI computing integrated machine built on the OpenAtom Dianhong (电鸿) IoT operating system for intelligent wind farm monitoring.\nKey technical details:\nArchitecture: A \u0026ldquo;vision, hearing, smell, sensing\u0026rdquo; (视、听、嗅、感) full-domain perception system deployed in wind farms Purpose: Addressing the long-standing pain points of traditional wind farm operations — high manual inspection costs, delayed hazard identification, and prominent operational safety risks Verification: The technology has been certified by the China Electricity Council (中国电力企业联合会) as reaching internationally advanced level Deployment: Scaled pilot validation completed in Jilin, Chongqing, and Jiangxi provinces\u0026rsquo; new energy stations Full-stack localization: The solution uses a fully domestic (国产化) technology chain, from hardware to software Institutional significance: The Datang Dianhong case represents a critical inflection point in the Dianhong open-source ecosystem — the transition from a single-SOE proof-of-concept to a multi-SOE institutional model.\nFirst, the replication of the Dianhong model across SOEs: The previous Dianhong case study (covered in the July 28 briefing) featured CR Power\u0026rsquo;s \u0026ldquo;RunDianHong\u0026rdquo; distribution. Now, barely two weeks later, a second major SOE — China Datang Group — has deployed its own Dianhong-based solution. This is not coincidental. The rapid succession of these two case studies suggests that the Dianhong community governance model is designed for replicability — the OpenAtom Foundation is systematically documenting and promoting SOE adoption patterns to create a template that other SOEs can follow.\nSecond, the institutional logic of \u0026ldquo;multi-modal AI + IoT OS\u0026rdquo;: The combination of multi-modal AI (vision, hearing, odor sensing, tactile sensing) with the Dianhong IoT OS represents a new institutional configuration. Rather than a generic IoT platform, the Dianhong ecosystem is evolving into a domain-specific AI infrastructure for the electric power industry. The AI capabilities are not bolted on as an afterthought but are integrated into the OS governance framework — the Dianhong community provides not just the OS kernel but also the AI toolchain, data standards, and deployment protocols. This is a significant institutional innovation: the sector-specific open-source community as an AI infrastructure organizer.\nThird, the role of the China Electricity Council certification: The certification by the China Electricity Council that the Datang Dianhong solution has reached \u0026ldquo;internationally advanced level\u0026rdquo; is an institutional signal. It means that the Dianhong open-source ecosystem has been formally recognized by the industry\u0026rsquo;s highest technical authority. This certification serves multiple purposes:\nLegitimacy: It validates the Dianhong model for other SOEs considering adoption Standardization: It creates a benchmark against which other Dianhong implementations can be measured Exportability: An internationally advanced certification from China\u0026rsquo;s electricity industry body could be used as a credential for international expansion, particularly in Belt and Road energy projects Fourth, the \u0026ldquo;smart sentry\u0026rdquo; (智慧哨兵) narrative: The article frames the Datang Dianhong solution as a \u0026ldquo;24-hour online smart sentry\u0026rdquo; — a metaphor that emphasizes continuous, autonomous, and unattended operation. This is consistent with the broader Chinese policy narrative of \u0026ldquo;unattended/substation-less\u0026rdquo; (无人化/少人化) transformation in critical infrastructure. The open-source nature of the Dianhong solution is not incidental to this narrative — it is essential, because only open-source can provide the transparency, auditability, and customization that critical infrastructure operators require when deploying AI systems in safety-critical environments.\nFifth, the strategic timing: The publication of this case study on July 28 — just two days after the CR Power RunDianHong case study was covered in this briefing series (July 28) and one day after the Kimi K3 article (July 27) — suggests a deliberate sequencing of institutional communications. The OpenAtom Foundation is building a narrative arc: first the general framework (Deep Report, July 15), then the sector-specific model (Dianhong, July 16), then the macro policy context (MIIT press conference, July 22), then the global AI narrative (Kimi K3, July 27), and now the concrete SOE deployment case (Datang Dianhong, July 28). Each article builds on the previous one, creating a cumulative case for the institutional viability of the State-Chartered Codebase model in critical infrastructure sectors.\nSource: OpenAtom Foundation Journalism\n🔍 WeChat Monitor OpenAtom Foundation Journalism:\n2026-07-28: \u0026ldquo;24小时在线\u0026rsquo;智慧哨兵\u0026rsquo;扎根风电场，开放原子电鸿激活风电智能新动能\u0026rdquo; — Datang Group Dianhong multi-modal AI for wind farm intelligent monitoring (this briefing) 2026-07-27: \u0026ldquo;中国开源大模型的\u0026rsquo;冲击\u0026rsquo;和启示\u0026rdquo; — Kimi K3 analysis from Guangming Daily (covered in July 29 briefing) 2026-07-23: \u0026ldquo;开放原子\u0026rsquo;园区行\u0026rsquo;香港站即将启幕\u0026rdquo; — openEuler Hong Kong launch preview (event held July 29, covered in July 29 briefing) 2026-07-22: MIIT press conference on open source ecosystem (covered in July 24 briefing) 2026-07-21: \u0026ldquo;70+ Policies Behind: Local Open Source Enters \u0026lsquo;Value Realization Period\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-20: \u0026ldquo;Embodied AI\u0026rsquo;s \u0026lsquo;Open Source Moment\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-16: \u0026ldquo;RunDianHong\u0026rdquo; — CR Power\u0026rsquo;s Dianhong IoT OS (covered in July 28 briefing) 2026-07-15: \u0026ldquo;China Open Source Deep Development Report (2025)\u0026rdquo; (covered in July 28 briefing) 🔍 Commentary The Dianhong Cascade: How One SOE Open-Source Deployment Becomes a Template for All\nToday\u0026rsquo;s briefing focuses on a single story — but one that carries significant institutional weight. The Datang Group Dianhong case study, published by the OpenAtom Foundation on July 28, reveals a pattern that is often missed in coverage of Chinese open source: the institutional machinery behind SOE open-source adoption.\n1. The Replication Mechanism\nThe CR Power RunDianHong case study (July 16) and the Datang Dianhong case study (July 28) are separated by only 12 days. This is not a coincidence. The OpenAtom Foundation is operating a replication mechanism — a systematic process of documenting, validating, and promoting SOE open-source deployments.\nThe replication mechanism works as follows:\nStep 1: An initial SOE (CR Power) develops a Dianhong-based solution for a specific domain (power generation IoT) Step 2: The OpenAtom Foundation publishes a case study, crediting the SOE\u0026rsquo;s innovation and framing the solution within the foundation\u0026rsquo;s governance narrative Step 3: Other SOEs (Datang Group) observe the first case study and develop their own Dianhong-based solutions, possibly with technical assistance from the foundation Step 4: The foundation publishes a second case study, demonstrating that the model is replicable Step 5: The collection of case studies builds a cumulative case for the Dianhong model, reducing adoption risk for subsequent SOEs This is a textbook example of institutional entrepreneurship — the OpenAtom Foundation is not just incubating open-source projects; it is actively creating the institutional conditions for their adoption by systematically documenting and publicizing successful deployments.\n2. The \u0026ldquo;Critical Infrastructure\u0026rdquo; Logic\nThe choice of domains for these case studies is strategically significant. Both CR Power and Datang Group are state-owned power generation enterprises — operators of critical national infrastructure. The Dianhong IoT OS is being deployed in power generation, transmission, and distribution — the most sensitive and security-critical sectors of the Chinese economy.\nFrom an institutional economics perspective, the Dianhong model represents a new governance form for critical infrastructure software. Rather than relying on proprietary software from foreign vendors (which creates supply chain risk) or purely community-governed open source (which may not meet regulatory requirements), the Dianhong model offers a third way: sector-specific open source governed by the OpenAtom Foundation, with direct participation from SOEs, and embedded within the regulatory framework of the electric power industry.\nThis model has implications beyond the power sector. If the Dianhong approach proves successful, it could be replicated in other critical infrastructure sectors — transportation, water, healthcare, telecommunications — each with its own sector-specific open-source community under the OpenAtom umbrella.\n3. The AI-OS Convergence\nThe Datang Dianhong case is particularly notable for its integration of multi-modal AI with the IoT OS. The \u0026ldquo;vision, hearing, smell, sensing\u0026rdquo; perception system is not a separate application running on top of the OS but is integrated into the Dianhong governance framework — the AI models, data pipelines, and deployment protocols are all part of the community\u0026rsquo;s standard toolchain.\nThis integration represents a convergence of two institutional logics: the open-source OS governance logic (version control, contribution models, lifecycle management) and the AI governance logic (model training data, inference protocols, safety validation). By combining these under a single community governance framework, the Dianhong model creates a vertically integrated AI infrastructure that is difficult to replicate in the Western open-source ecosystem, where OS foundations and AI governance typically operate in separate institutional spheres.\n4. The Certification as Institutional Signal\nThe China Electricity Council\u0026rsquo;s certification of the Datang Dianhong solution as \u0026ldquo;internationally advanced\u0026rdquo; is more than a technical endorsement — it is an institutional signal to the entire Chinese power industry. The signal says: \u0026ldquo;This open-source solution has been validated by the industry\u0026rsquo;s highest authority. Adoption is safe, legitimate, and encouraged.\u0026rdquo;\nThis certification interacts with the Chinese regulatory environment in a specific way. China\u0026rsquo;s Critical Information Infrastructure (CII) regulations require operators to use software that meets certain security and reliability standards. By obtaining an industry-level certification, the Dianhong ecosystem provides a compliance pathway for SOEs that want to adopt open-source solutions without violating regulatory requirements. This is a crucial institutional function — one that Western open-source foundations do not typically perform, because they operate in regulatory environments where industry certification is not a prerequisite for open-source adoption.\n5. What This Means for the Global Open-Source Ecosystem\nThe Dianhong cascade has implications for the global open-source ecosystem:\nA new model for SOE open-source adoption: The Dianhong model demonstrates that SOEs can be active producers of open-source software, not just consumers. This is a significant departure from the Western pattern, where open-source production is dominated by tech companies and individual developers.\nSector-specific open-source governance: The Dianhong model creates a template for sector-specific open-source communities that are integrated with industry regulation. This is a model that could be adopted by other countries seeking to develop open-source solutions for critical infrastructure sectors.\nThe AI-OS convergence as a governance challenge: The integration of AI into the Dianhong OS governance framework raises questions about how AI governance should be handled in open-source communities. The Dianhong model\u0026rsquo;s approach — embedding AI within the OS governance framework rather than treating it as a separate concern — is one possible answer, but it may not be the right answer for all contexts.\nThe replication mechanism as a governance strategy: The OpenAtom Foundation\u0026rsquo;s systematic documentation and promotion of SOE case studies is a governance strategy that deserves attention. By creating a public record of successful deployments, the foundation reduces the institutional barriers to open-source adoption and builds a cumulative case for its governance model. This is a strategy that other open-source foundations could learn from.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-30/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-30\"\u003eChina Open Source Daily — 2026-07-30\u003c/h2\u003e\n\u003ch3 id=\"-institutional-change-soe-open-source-ecosystem-expansion\"\u003e🏗️ Institutional Change: SOE Open-Source Ecosystem Expansion\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. Datang Group Dianhong: Multi-Modal AI Computing for Wind Farm Intelligent Monitoring\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 28, the OpenAtom Foundation published a case study of \u003cstrong\u003eChina Datang Group Digital Technology (大唐数科, \u0026ldquo;Datang Digital Tech\u0026rdquo;)\u003c/strong\u003e, a subsidiary of \u003cstrong\u003eChina Datang Group (中国大唐集团)\u003c/strong\u003e — one of China\u0026rsquo;s \u0026ldquo;Big Five\u0026rdquo; state-owned power generation enterprises. The company developed a \u003cstrong\u003emulti-modal AI computing integrated machine\u003c/strong\u003e built on the \u003cstrong\u003eOpenAtom Dianhong (电鸿) IoT operating system\u003c/strong\u003e for intelligent wind farm monitoring.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-30"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-29 🏛️ Policy \u0026amp; Global Governance 1. Kimi K3: Chinese Open-Source Large Model Challenges Closed-Source AI Dominance\nOn July 27, the OpenAtom Foundation published a Guangming Daily article titled \u0026ldquo;The \u0026lsquo;Impact\u0026rsquo; and Insights of Chinese Open-Source Large Models\u0026rdquo; (中国开源大模型的\u0026quot;冲击\u0026quot;和启示), analyzing the global significance of Moonshot AI\u0026rsquo;s (月之暗面) Kimi K3 open-source model.\nKey technical details:\nScale: 2.8 trillion parameters, using a highly efficient Mixture-of-Experts (MoE) architecture Performance: Topped the Frontend Code Arena benchmark with 1679 points, surpassing Claude Fable 5 (1631), GPT-5.6 Sol (1618), and GLM-5.2 Max (1587) — the first open-source model ever to surpass all closed-source models on this benchmark Efficiency: Activate only a fraction of expert modules per inference, with self-developed computation mechanisms achieving comparable or superior performance at a fraction of the energy consumption of top closed-source models Pricing: API priced at just one-third of Claude Fable 5, with comparable coding capability Openness: Full model weights to be open-sourced by end of July 2026, enabling local deployment and sovereign control Global reactions:\nElon Musk described the achievement as \u0026ldquo;impressive\u0026rdquo; Jensen Huang (NVIDIA CEO) told media that Chinese open-source AI models are \u0026ldquo;very excellent\u0026rdquo; and that US enterprises should use them, warning that restricting open-source models could \u0026ldquo;weaken America\u0026rsquo;s competitiveness in AI\u0026rdquo; Vercel CEO Guillermo Rauch stated after testing that Kimi K3 \u0026ldquo;leads Fable 5 in real-world Web engineering delivery\u0026rdquo; El País (Spain) sharply questioned: \u0026ldquo;How can Anthropic now defend its pricing?\u0026rdquo; Bloomberg described Kimi K3 as \u0026ldquo;shaking up global tech markets\u0026rdquo; Nikkei analyzed that the closed-source model of Anthropic and OpenAI depends on a market oligopoly that Kimi K3 potentially disrupts Institutional significance: The Kimi K3 moment represents a paradigm shift in the institutional economics of AI. The article frames the development around four axes:\nFirst, the \u0026ldquo;efficiency path\u0026rdquo; vs. the \u0026ldquo;scaling path\u0026rdquo; : Closed-source AI development follows a logic of infinite expansion — larger parameters, more chips, higher energy consumption. Kimi K3 demonstrates that architectural innovation, not brute-force compute scaling, can achieve equivalent intelligence density. This is not merely a technical divergence but a fundamental institutional choice about resource allocation — one that is vastly more accessible to the Global South, where power infrastructure and compute budgets are constrained.\nSecond, the \u0026ldquo;open-source disruption of monopoly pricing\u0026rdquo; : The closed-source AI market has functioned as a textbook oligopoly, with a handful of companies (Anthropic, OpenAI) maintaining premium pricing on the assumption that no competitor can match their capability. Kimi K3\u0026rsquo;s API pricing at one-third of Claude Fable 5, with full-weight open-sourcing, fundamentally breaks this pricing power. From an institutional economics perspective, this is a credible commitment to competition — the open-source model cannot be withdrawn or re-priced, creating a structural constraint on the oligopoly\u0026rsquo;s pricing behavior.\nThird, the 2026 WAIC and the institutionalization of open-source AI governance: The article prominently features the 2026 World AI Conference (WAIC) outcomes, where 29 countries signed the Agreement on Establishing the World AI Cooperation Organization in Shanghai. The WAIC Chairman\u0026rsquo;s Statement enshrined \u0026ldquo;open source and openness\u0026rdquo; as a key path for AI inclusive development, stating that the international community \u0026ldquo;should responsibly encourage the co-building of open-source ecosystems and actively create open and inclusive international open-source communities.\u0026rdquo; China committed to providing 5,000 AI training slots for developing countries over the next five years and establishing international AI application cooperation centers with ASEAN, the Arab League, the African Union, CELAC, the SCO, and BRICS. This represents a multilateral institutional framework for open-source AI governance — a direct alternative to the unilateral technology blockade model.\nFourth, the \u0026ldquo;smart divide\u0026rdquo; framing: The article argues that the greatest risk in AI global governance is the \u0026ldquo;smart divide\u0026rdquo; — the monopolization of AI benefits by a few nations and corporations. Kimi K3\u0026rsquo;s open-source, low-cost, energy-efficient model is positioned as a bridge across this divide, enabling developing countries to deploy advanced AI capabilities without surrendering data sovereignty or innovation capacity to foreign cloud providers. This is a new institutional narrative that frames open-source AI as a global public good rather than a commercial product.\nSource: OpenAtom Foundation Journalism — Guangming Daily\n🏗️ Institutional Change 2. OpenAtom \u0026ldquo;Park Tour\u0026rdquo; Hong Kong Station: openEuler\u0026rsquo;s First Overseas Regional Community\nOn July 23, the OpenAtom Foundation announced the \u0026ldquo;Park Tour\u0026rdquo; (园区行) Hong Kong Station — the Open Source Ecosystem Internationalization and openEuler Hong Kong Launch event, taking place today (July 29) at the Hyatt Regency Sha Tin, Hong Kong. The event is hosted by the OpenAtom Foundation, co-organized by the Hong Kong Logistics and Supply Chain MultiTech R\u0026amp;D Centre (LSCM) and the openEuler Community.\nKey institutional developments:\nopenEuler Hong Kong User Group officially established: This is the first official regional community established outside mainland China by the OpenAtom Foundation. Its organizational model and operational mechanisms will serve as a reference for the internationalization of future open-source projects under the foundation\u0026rsquo;s governance framework.\nHong Kong industry use cases presented: The event showcases openEuler deployment cases across Hong Kong\u0026rsquo;s Customs and Excise Department, Police Force, Hong Kong Exchanges and Clearing (HKEX), Hospital Authority, and LSCM — covering public governance, financial services, and healthcare sectors. These are significant proof points for openEuler\u0026rsquo;s real-world applicability in a mature, internationally-oriented economy.\nGovernment-industry-academia高层齐聚: The event brings together senior representatives from government, industry, academia, and research institutions to discuss how open-source technology can better integrate locally while connecting globally. Topics include kernel technology and AI supply chain security.\nInstitutional significance: The Hong Kong Park Tour represents a critical inflection point in the OpenAtom Foundation\u0026rsquo;s internationalization strategy. Several dimensions deserve analysis:\nFirst, Hong Kong as an institutional bridgehead: Hong Kong\u0026rsquo;s unique position — a Special Administrative Region of China with a common law legal system, international financial infrastructure, and deep integration with global technology markets — makes it an ideal institutional laboratory for testing the exportability of China\u0026rsquo;s open-source governance model. The OpenAtom Foundation is not merely promoting openEuler in Hong Kong; it is using Hong Kong as a proof of concept for the translatability of State-Chartered Codebase governance frameworks into international contexts.\nSecond, the sector-specific use case approach: The selection of Hong Kong Customs, Police, HKEX, and Hospital Authority as deployment cases is strategically significant. These are high-trust, high-regulation sectors where open-source adoption faces the highest institutional barriers. Successful deployment in these sectors signals that openEuler can meet the rigorous security, compliance, and reliability requirements of regulated industries — a credential that is essential for international expansion.\nThird, the institutional isomorphism of regional communities: The establishment of the Hong Kong User Group as the first overseas regional community creates an institutional template that can be replicated. The organizational model, governance mechanisms, and operational processes developed for Hong Kong will likely be adapted for other international markets. This is a classic case of institutional isomorphism — the creation of a standardized organizational form that can be transferred across jurisdictions.\nFourth, timing and strategic context: The event\u0026rsquo;s timing — July 29, 2026 — coincides with the aftermath of the US export control escalation (July 22-23 covered in the July 24 briefing) and the Kimi K3 global attention wave. The openEuler Hong Kong launch can be read as a strategic signal that China\u0026rsquo;s open-source infrastructure is not retreating from internationalization in response to sanctions but is instead accelerating its overseas institutional presence. Hong Kong serves as a jurisdictional hedge — a location where open-source governance can operate under a different legal and regulatory framework than mainland China, potentially reducing friction with international partners.\nSource: OpenAtom Foundation Journalism\n🔍 WeChat Monitor OpenAtom Foundation Journalism:\n2026-07-27: \u0026ldquo;The \u0026lsquo;Impact\u0026rsquo; and Insights of Chinese Open-Source Large Models\u0026rdquo; — Kimi K3 analysis from Guangming Daily 2026-07-23: \u0026ldquo;OpenAtom \u0026lsquo;Park Tour\u0026rsquo; Hong Kong Station即将启幕\u0026rdquo; — openEuler Hong Kong launch preview 2026-07-22: MIIT press conference on open source ecosystem (covered in July 24 briefing) 2026-07-21: \u0026ldquo;70+ Policies Behind: Local Open Source Enters \u0026lsquo;Value Realization Period\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-20: \u0026ldquo;Embodied AI\u0026rsquo;s \u0026lsquo;Open Source Moment\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-16: \u0026ldquo;RunDianHong\u0026rdquo; — CR Power\u0026rsquo;s Dianhong IoT OS (covered in July 28 briefing) 2026-07-15: \u0026ldquo;China Open Source Deep Development Report (2025)\u0026rdquo; (covered in July 28 briefing) 🔍 Commentary The July 29 Turning Point: Two Faces of Chinese Open-Source Internationalization\nToday\u0026rsquo;s briefing covers two stories that, taken together, reveal the dual strategy of Chinese open-source internationalization in the post-Kimi K3 era.\n1. The Kimi K3 Moment: Open-Source AI as a Global Public Good\nThe Kimi K3 article published by the OpenAtom Foundation represents a significant framing exercise. By republishing a Guangming Daily article (the official newspaper of the CPC Central Committee) on the foundation\u0026rsquo;s journalism platform, the OpenAtom Foundation is signaling institutional alignment between the foundation\u0026rsquo;s perspective and the party-state\u0026rsquo;s official media narrative.\nThe article\u0026rsquo;s framing of Kimi K3 as a \u0026ldquo;bridge across the smart divide\u0026rdquo; — rather than merely a commercial product — is a deliberate institutional choice. It positions Chinese open-source AI not as a competitor in a market but as a contributor to a global public good. This framing has several consequences:\nLegitimacy: By invoking the \u0026ldquo;29 countries\u0026rdquo; WAIC agreement and the \u0026ldquo;5000 training slots\u0026rdquo; commitment, the article frames Chinese open-source AI within a multilateral governance framework, not a unilateral national strategy. Reciprocity: The implicit argument is that if Chinese open-source AI is a global public good, then restrictions on its development (export controls, technology blockades) are not just commercial disputes but harm to the global commons. Institutional entrepreneurship: The OpenAtom Foundation, by publishing and promoting this article, positions itself as the institutional voice of the open-source AI global public good — a role that enhances its relevance in international AI governance discussions. 2. The Hong Kong Bridgehead: Institutional Export of the State-Chartered Codebase Model\nThe openEuler Hong Kong launch is the institutional counterpart to the Kimi K3 narrative. While Kimi K3 represents the product-level internationalization of Chinese open-source AI, the Hong Kong Park Tour represents the governance-level internationalization of the Chinese open-source model.\nThe openEuler Hong Kong User Group is significant not because of its technical specifications (openEuler has been available internationally for years) but because it represents the replication of the OpenAtom Foundation\u0026rsquo;s governance model in a jurisdiction outside mainland China. The organizational template — a regional user group under the foundation\u0026rsquo;s umbrella, with localized governance, industry use cases, and government-academia partnerships — is designed to be scalable and transferable.\n3. The Strategic Complementarity\nThese two developments — the Kimi K3 narrative and the openEuler Hong Kong launch — are strategically complementary:\nKimi K3 provides the \u0026ldquo;why\u0026rdquo; : Chinese open-source AI is valuable, efficient, and globally beneficial Hong Kong openEuler provides the \u0026ldquo;how\u0026rdquo; : China\u0026rsquo;s open-source governance infrastructure can operate in international contexts Together, they constitute a comprehensive institutional response to the export control regime: rather than retreating from globalization, Chinese open-source institutions are simultaneously proving their technical value (Kimi K3) and expanding their governance footprint (Hong Kong openEuler).\n4. Institutional Implications for the Global Open-Source Ecosystem\nThe dual strategy carries implications for the global open-source ecosystem:\nCompeting governance models: The OpenAtom Foundation\u0026rsquo;s international expansion introduces a governance model that differs significantly from Western foundations (Apache, Linux Foundation). The OpenAtom model is more closely integrated with state industrial policy, sector-specific regulation, and geopolitical strategy. As it expands internationally, it will create institutional competition — not just in technology but in governance philosophy.\nThe jurisdictional question: The Hong Kong experiment raises an important institutional question: can the State-Chartered Codebase governance model operate effectively under a common law legal system? If successful, it would demonstrate that the model is not dependent on China\u0026rsquo;s civil law framework, potentially expanding its appeal in other jurisdictions.\nThe \u0026ldquo;open-source public good\u0026rdquo; narrative as soft power: The Kimi K3 framing as a global public good represents a new form of technological soft power. By combining technical excellence (benchmark-topping performance) with inclusive governance (open-source, affordable pricing, global deployment capability), China is constructing a narrative that challenges the \u0026ldquo;technology as competitive advantage\u0026rdquo; logic of the US-led AI ecosystem.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-29/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-29\"\u003eChina Open Source Daily — 2026-07-29\u003c/h2\u003e\n\u003ch3 id=\"-policy--global-governance\"\u003e🏛️ Policy \u0026amp; Global Governance\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. Kimi K3: Chinese Open-Source Large Model Challenges Closed-Source AI Dominance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 27, the OpenAtom Foundation published a Guangming Daily article titled \u003cstrong\u003e\u0026ldquo;The \u0026lsquo;Impact\u0026rsquo; and Insights of Chinese Open-Source Large Models\u0026rdquo;\u003c/strong\u003e (中国开源大模型的\u0026quot;冲击\u0026quot;和启示), analyzing the global significance of Moonshot AI\u0026rsquo;s (月之暗面) \u003cstrong\u003eKimi K3\u003c/strong\u003e open-source model.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-29"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-28 🏛️ Institutional Change 1. CCF ChinaOSC 2026 Preview: Xenomai Sub-Forum on Real-Time OS for Embodied Intelligence, Aerospace, and Industrial Control\nOn July 25, the CCF Open Source Development Technology Committee (CCF ODTC) published a detailed preview of the 2026 CCF China Open Source Conference (ChinaOSC), scheduled for August 15-16 in Chongqing. The conference, jointly organized by CCF and Chongqing University, is themed \u0026ldquo;Meet Open Source in Chongqing, Digital Intelligence Opens New Horizons\u0026rdquo; (渝见开源，数智启新).\nA featured sub-forum highlighted in the preview is \u0026ldquo;Open Source Real-Time, Embodied Symbiosis — Xenomai for Embodied Intelligence, Aerospace Computing, Industrial Control, and Smart Vehicles.\u0026rdquo; Key details of the forum:\nForum chair: Dr. Wang Guangfeng (王广锋), founder and chair of the Xenomai China Community, GM of Beijing Nailian Technology, director of the AI+Industrial OS Joint Lab, and PMC member of the OpenAtom Foundation\u0026rsquo;s M-Robots community Core thesis: As intelligent systems move from virtual simulation to physical interaction, deterministic, low-latency, and reliable execution in perception-planning-control loops becomes critical Key question: How to bridge the gap between intelligent decision-making and deterministic execution? How can Xenomai balance Linux\u0026rsquo;s open ecosystem with real-time, reliability, and security requirements for critical systems? Application domains: Embodied intelligence robots, aerospace computing platforms, industrial control systems, and intelligent vehicle platforms The conference is positioned as a high-level academic-industry gathering, with sessions covering AI for Software Engineering (AI4SE), Software Engineering for AI (SE4AI), open source code generation and trusted maintenance, open source operating systems, industrial software, open source LLMs and AI agent satellites, open source scientific computing, and open source supply chain security.\nInstitutional significance: The CCF ChinaOSC has become the premier annual gathering for China\u0026rsquo;s open source research community. The Xenomai sub-forum is particularly interesting because it bridges the gap between academic open source (Xenomai\u0026rsquo;s dual-kernel real-time architecture) and industrial applications (embodied AI, aerospace, automotive). The participation of the OpenAtom Foundation\u0026rsquo;s M-Robots PMC member as forum chair signals the increasing convergence between the foundation\u0026rsquo;s project incubation system and the academic research community — a trend that could accelerate the transition of research-stage open source projects into the foundation\u0026rsquo;s governance framework.\nSource: CCF ODTC WeChat Article\n📊 Research \u0026amp; Analysis 2. OpenAtom Deep Report: Open Source AI Enters \u0026ldquo;Ecosystem Decisive\u0026rdquo; Era\nOn July 15, the OpenAtom Foundation published the \u0026ldquo;China Open Source Deep Development Report (2025)\u0026rdquo; (《中国开源发展深度报告（2025）》), a comprehensive analysis of the Chinese open source landscape. The report\u0026rsquo;s central thesis: open source AI has entered an \u0026ldquo;ecosystem decisive\u0026rdquo; (生态决胜) era.\nKey findings of the report:\nDeepSeek as inflection point: 2025 marked a watershed moment, with DeepSeek\u0026rsquo;s open-source models triggering global attention and accelerating the convergence of open source and AI Expanding scope of openness: The report identifies that open source has expanded from code sharing to encompass models, inference frameworks, agents, development platforms, data toolchains, and AI governance frameworks AI as infrastructure: Open source is no longer just a code-sharing methodology but has become infrastructure for organizing AI innovation, accelerating technical iteration, and driving industrial coordination Ecosystem competition: The report argues that the competitive landscape for AI has shifted from model capability to ecosystem depth — the ability to attract developers, support toolchains, and integrate with industrial applications Institutional significance: This report represents the OpenAtom Foundation\u0026rsquo;s most comprehensive attempt to frame the AI-open source convergence. The \u0026ldquo;ecosystem decisive\u0026rdquo; framing is significant because it shifts the policy conversation from \u0026ldquo;how many models?\u0026rdquo; to \u0026ldquo;how deep is the ecosystem?\u0026rdquo; — a framing that aligns with the foundation\u0026rsquo;s institutional role as an ecosystem organizer rather than a model developer. The report\u0026rsquo;s publication date (July 15) — just before the MIIT press conference (July 20) and the White House sanctions escalation (July 22-23) — suggests it was timed to establish a baseline for the public debate that followed.\nSource: OpenAtom Foundation Journalism\n3. OpenAtom Analysis: AI-Driven Open Source Industry Transformation\nOn July 13, the OpenAtom Foundation published a companion analysis, \u0026ldquo;AI Drives All-Domain Transformation of the Open Source Industry\u0026rdquo; (AI驱动开源产业全域变革), arguing that AI is fundamentally redefining what open source means.\nKey arguments:\nBeyond source code: The analysis argues that the object of openness has expanded from software source code to models, data, computing power, agents, terminals, business models, and governance rules New institutional requirements: The expanded scope of openness creates new requirements for governance frameworks, legal instruments, and business models that the current open source infrastructure is not yet equipped to handle Structural tension: The analysis identifies a tension between the rapid pace of AI-driven openness and the slower pace of institutional adaptation — a classic institutional economics problem of institutional lag Institutional significance: This analysis, published alongside the Deep Report, serves as a policy framing document. By explicitly identifying the gap between technical openness and institutional readiness, the OpenAtom Foundation positions itself as the necessary intermediary — the institution that can bridge the gap between rapid technical change and the slower pace of governance adaptation. This is a sophisticated institutional strategy: identify a problem that only your organization can solve.\nSource: OpenAtom Foundation Journalism\n🔧 Open Source in Practice 4. OpenAtom Dianhong: CR Power\u0026rsquo;s \u0026ldquo;RunDianHong\u0026rdquo; IoT OS for Power Systems\nOn July 16, the OpenAtom Foundation published a case study of \u0026ldquo;RunDianHong\u0026rdquo; (润电鸿), the first enterprise distribution of the OpenAtom Dianhong (电鸿) community — an open source IoT operating system for electric power systems developed by China Resources Power (华润电力).\nKey technical details:\nPurpose: A generation-side IoT OS digital base for new energy power stations (wind, solar) Architecture: Full-stack \u0026ldquo;cloud-management-edge-end\u0026rdquo; (云管边端) connectivity Deployment: Scaled pilot validation in Shandong province new energy stations Functionality: Smart inspection solutions for new energy stations, addressing fragmentation of multi-terminal data standards Institutional significance: The Dianhong open source community represents a sector-specific open source governance model — an open source foundation organized around a specific industrial domain (electric power) rather than a general-purpose technology stack. This is a distinctly Chinese institutional innovation: the OpenAtom Foundation incubates both general-purpose projects (OpenHarmony, openEuler) and sector-specific communities (Dianhong for electric power, M-Robots for robotics, OpenLET for embodied intelligence). The sector-specific model allows the foundation to embed open source governance directly into industrial policy frameworks, creating a vertical integration of open source and industrial regulation.\nThe RunDianHong case is particularly notable because it involves a state-owned enterprise (CR Power) developing and open-sourcing an industrial IoT OS — a pattern that is becoming increasingly common in China\u0026rsquo;s energy sector. This represents a new institutional logic: SOEs are not just adopting open source but actively producing it, often as a strategic tool for supply chain standardization and ecosystem control.\nSource: OpenAtom Foundation Journalism\n🔍 WeChat Monitor CCF Open Source Development Technology Committee (CCF ODTC):\n2026-07-25: \u0026ldquo;2026 CCF China Open Source Conference Sub-Forum Introduction — Open Source Real-Time, Embodied Symbiosis: Xenomai for Embodied Intelligence, Aerospace Computing, Industrial Control, and Smart Vehicles\u0026rdquo; — detailed preview of the August 15-16 conference OpenAtom Foundation Journalism:\n2026-07-22: MIIT press conference on open source ecosystem (covered in July 24 briefing) 2026-07-21: \u0026ldquo;70+ Policies Behind: Local Open Source Enters \u0026lsquo;Value Realization Period\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-20: \u0026ldquo;Embodied AI\u0026rsquo;s \u0026lsquo;Open Source Moment\u0026rsquo;\u0026rdquo; (covered in July 27 briefing) 2026-07-16: \u0026ldquo;RunDianHong\u0026rdquo; — CR Power\u0026rsquo;s Dianhong IoT OS for power systems 2026-07-15: \u0026ldquo;China Open Source Deep Development Report (2025)\u0026rdquo; — Open Source AI enters \u0026ldquo;ecosystem decisive\u0026rdquo; era 2026-07-13: \u0026ldquo;AI Drives All-Domain Transformation of the Open Source Industry\u0026rdquo; 🔍 Commentary The Week of Institutional Documentation: Three Developments in Chinese Open Source Governance\nThis briefing covers several articles that were published between July 13-25 but not yet reported in this daily series. Together, they reveal a pattern of institutional documentation — the OpenAtom Foundation and CCF are systematically building a public record of China\u0026rsquo;s open source institutional development.\n1. The CCF ChinaOSC as a Research-Policy Bridge\nThe CCF China Open Source Conference has evolved from a purely academic gathering into a research-policy bridge institution. The Xenomai sub-forum exemplifies this: it brings together academic researchers (Xenomai is a research project originally from the French INRIA), Chinese community builders (the Xenomai China Community), and the OpenAtom Foundation\u0026rsquo;s project incubation system (M-Robots community). This multi-institutional collaboration is a governance structure that has no direct equivalent in the Western open source ecosystem, where academic research, community development, and foundation governance typically operate in separate institutional spheres.\nThe conference\u0026rsquo;s focus on AI4SE (AI for Software Engineering) and SE4AI (Software Engineering for AI) as core themes is also institutionally significant. It signals that the Chinese open source research community is treating the AI-software relationship as a two-way street — not just AI as a tool for software development, but software engineering as a discipline for managing AI systems. This framing has implications for how open source governance is designed: if AI systems are themselves software that must be engineered, then existing open source governance frameworks (licensing, contribution models, lifecycle management) can be adapted rather than reinvented.\n2. The Deep Report as a Policy Framing Document\nThe \u0026ldquo;ecosystem decisive\u0026rdquo; framing in the OpenAtom Deep Report is a deliberate policy intervention. By shifting the focus from model capability to ecosystem depth, the report creates a new metric for evaluating open source AI success — one that the OpenAtom Foundation is uniquely positioned to measure and influence. The report\u0026rsquo;s publication just before the MIIT press conference (which provided the first official quantification of the open source ecosystem) suggests strategic coordination between the foundation and the ministry.\nThe report\u0026rsquo;s institutional logic is clear: if the decisive factor in AI open source is ecosystem depth, and the foundation is the primary ecosystem organizer, then the foundation\u0026rsquo;s institutional importance is structurally guaranteed. This is a classic example of what institutional economists call \u0026ldquo;institutional entrepreneurship\u0026rdquo; — the creation of new metrics and categories that enhance the relevance of the institution that creates them.\n3. The Dianhong Model: Sector-Specific Open Source Governance\nThe CR Power RunDianHong case study represents a significant institutional innovation: sector-specific open source communities within the broader OpenAtom governance framework. The Dianhong model embeds open source governance directly into the regulatory structure of the electric power industry, creating a vertical integration of open source and industrial policy.\nThis model has several implications:\nSOEs as open source producers: State-owned enterprises are becoming active producers of open source software, not just consumers. This changes the incentive structure of Chinese open source, as SOEs bring different motivations (regulatory compliance, ecosystem control, supply chain standardization) than tech companies or individual developers. Industrial regulation through open source: By governing the Dianhong community, the OpenAtom Foundation effectively becomes a participant in the regulatory structure of the electric power industry. This is a role that Western open source foundations do not play. Scalability of the model: If the Dianhong model proves successful, it could be replicated in other regulated industries (transportation, healthcare, manufacturing), creating a network of sector-specific open source communities under the OpenAtom umbrella. 4. The Institutional Documentation Function\nPerhaps the most notable pattern across these articles is the institutional documentation function that the OpenAtom Foundation and CCF are performing. Almost every article is explicitly framed as a case study, a report, an analysis, or a preview — not just news, but documentation of institutional practice. This is a governance strategy in itself: by creating a public record of how open source institutions work in China, the foundation and the CCF are simultaneously:\nLegitimizing their institutional roles through documented activity Educating stakeholders (policymakers, enterprises, developers) about the institutional infrastructure Creating precedents that can be cited in future policy discussions Building a historical record that can be used to measure institutional evolution This documentation function is itself an institutional innovation — one that is particularly important in a context where the institutional infrastructure of open source is still being built and needs to be visible to be credible.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-28/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-28\"\u003eChina Open Source Daily — 2026-07-28\u003c/h2\u003e\n\u003ch3 id=\"-institutional-change\"\u003e🏛️ Institutional Change\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. CCF ChinaOSC 2026 Preview: Xenomai Sub-Forum on Real-Time OS for Embodied Intelligence, Aerospace, and Industrial Control\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 25, the \u003cstrong\u003eCCF Open Source Development Technology Committee (CCF ODTC)\u003c/strong\u003e published a detailed preview of the \u003cstrong\u003e2026 CCF China Open Source Conference (ChinaOSC)\u003c/strong\u003e, scheduled for \u003cstrong\u003eAugust 15-16 in Chongqing\u003c/strong\u003e. The conference, jointly organized by CCF and Chongqing University, is themed \u003cstrong\u003e\u0026ldquo;Meet Open Source in Chongqing, Digital Intelligence Opens New Horizons\u0026rdquo;\u003c/strong\u003e (渝见开源，数智启新).\u003c/p\u003e","title":"China Open Source Daily — 2026-07-28"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-27 🏛️ Policy \u0026amp; Regulation 1. 70+ Local Open Source Policies: China\u0026rsquo;s Local Open Source Enters \u0026ldquo;Value Realization Period\u0026rdquo;\nThe OpenAtom Foundation published a comprehensive analysis on July 21 documenting the evolution of China\u0026rsquo;s local open source policy landscape. According to the analysis, the number of local policies explicitly containing \u0026ldquo;open source\u0026rdquo; content has exceeded 70 nationwide, marking a critical inflection point for the institutionalization of open source at the subnational level.\nKey milestones:\nBeijing and Shanghai have released dedicated open source system construction documents Zhejiang, Jiangsu, Guangdong, Guizhou, and Chongqing are embedding open source into AI, advanced manufacturing, intelligent connected vehicles, and embodied AI strategies The analysis identifies four critical shifts reshaping local open source development logic:\nShift 1: From \u0026ldquo;standalone open source policies\u0026rdquo; to \u0026ldquo;embedded into industrial systems\u0026rdquo; Open source is no longer treated as a separate policy category. Instead, it is being directly embedded into local industrial development frameworks. Beijing connects open source infrastructure to key industrial布局; Shanghai coordinates open source systems with intelligent computing cloud industry development; Zhejiang uses open source LLM ecosystems as a key lever for AI innovation. The formulation has shifted from \u0026ldquo;supporting open source\u0026rdquo; to \u0026ldquo;using open source to develop industry.\u0026rdquo;\nShift 2: From \u0026ldquo;code openness\u0026rdquo; to \u0026ldquo;full-element openness\u0026rdquo; AI is redefining the boundaries of open source. The object of openness has expanded from software source code to models, data, development frameworks, toolchains, computing resources, and hardware designs — a new paradigm of full-element collaborative openness. Local competition has upgraded from project counts and code volume to competition in resource allocation, technical coordination, and ecosystem organization capabilities.\nShift 3: From \u0026ldquo;online communities\u0026rdquo; to \u0026ldquo;dual online + offline centers\u0026rdquo; Local open source ecosystems are transitioning to a dual-center model: online communities aggregate innovation resources, while offline venues (open source parks, joint labs, developer centers, testing/certification centers) serve as industrial conversion nodes. Online communities determine whether resources can be gathered; offline venues determine whether outcomes can be validated, incubated, and scaled.\nShift 4: From \u0026ldquo;general technology supply\u0026rdquo; to \u0026ldquo;industrial scenario empowerment\u0026rdquo; The measure of local open source capability is shifting from platform/project/developer counts to whether open source can actually enter industrial processes and solve real problems. Open source operating systems, LLMs, and RISC-V are entering transportation, manufacturing, government services, healthcare, intelligent connected vehicles, and embodied AI scenarios.\nThree structural contradictions that must be resolved:\nSupply convergence vs. demand differentiation: Incubation platforms, developer conferences, and industry funds have become standard configurations, but precise alignment with local dominant industries remains weak Construction investment vs. professional operations: Platforms can be built with funding; communities require professional governance, sustained operations, and developer trust. The risk of \u0026ldquo;built but cannot run\u0026rdquo; (建得起、转不动) is real Traditional talent evaluation vs. open source contribution logic: Code commits, vulnerability fixes, documentation, and community maintenance — long-term, distributed contributions — are not yet recognized by talent evaluation systems for professional titles, talent plans, or science and technology awards Five upgrades for the 15th Five-Year Plan period:\nFrom \u0026ldquo;copying templates\u0026rdquo; to local adaptation based on local industry pain points From \u0026ldquo;project construction\u0026rdquo; to full lifecycle service (incubation → community operations → legal compliance → supply chain security → commercial conversion) From \u0026ldquo;talent quantity\u0026rdquo; to \u0026ldquo;contribution = value\u0026rdquo; — integrating major open source contributions into professional title evaluation and talent plans From \u0026ldquo;individual布局\u0026rdquo; to regional coordination via Jing-Jin-Ji, Yangtze River Delta, and Greater Bay Area From \u0026ldquo;application followership\u0026rdquo; to root technology breakthroughs — sustained support for OpenHarmony, openEuler, LLMs, AI frameworks, embodied AI, AI chips, and intelligent connected vehicles Source: OpenAtom Foundation Journalism\n2. Zhejiang Province Issues 7-Department Open Source System Construction Implementation Plan\nAs the most concrete example of the \u0026ldquo;value realization\u0026rdquo; trend, Zhejiang Province has become the latest locality to publish a comprehensive open source system policy. Seven provincial departments jointly issued the Implementation Opinions on Accelerating the Construction of Zhejiang Province\u0026rsquo;s Open Source System (《关于加快推进浙江省开源体系建设的实施意见》).\nKey features of the Zhejiang plan:\nPolicy scope: Covers infrastructure, technology supply, industrial application, entity cultivation, talent evaluation, security governance, and commercial conversion Expanded openness: The object of openness extends from code to hardware, data, models, and standards Policy instruments: Open source contribution evaluation, software bill of materials (SBOM), project whitelist, computing vouchers (算力券), data vouchers (数据券), and government procurement 2027 targets: 50+ high-quality open source projects, 150+ open source innovation enterprises, 100+ application benchmarks Institutional significance: The Zhejiang plan represents a qualitative leap in policy logic — from principle advocacy to institutional supply, from fragmented support to full-chain layout, from supporting individual projects to cultivating complete ecosystems, and from \u0026ldquo;supporting open source\u0026rdquo; to \u0026ldquo;creating industrial value through open source.\u0026rdquo;\nSource: OpenAtom Foundation Journalism\n🤖 Open Source \u0026amp; AI 3. Embodied AI\u0026rsquo;s \u0026ldquo;Open Source Moment\u0026rdquo; — OpenLET Dataset Surpasses 3 Million Downloads\nThe OpenAtom Foundation published a detailed feature on July 20 analyzing the rapid open sourcing of embodied AI (具身智能) technology. The piece, authored by Zhang Nan from Software and Integrated Circuits journal, argues that a clear pattern is emerging: when a technology transitions from frontier exploration by a few to industrial application by many, open source is not a luxury — it is the necessary path.\nFrom DeepSeek to OpenLET: Following DeepSeek\u0026rsquo;s \u0026ldquo;AI open source moment,\u0026rdquo; embodied AI\u0026rsquo;s \u0026ldquo;open source moment\u0026rdquo; is now unfolding through the OpenLET community, an open source embodied intelligence platform hosted under the OpenAtom Foundation.\nKey developments:\nData openness — the critical bottleneck:\nUnitree Robotics (宇树科技) has open-sourced TB-level humanoid robot datasets on its official website, described as \u0026ldquo;currently the most comprehensive open-source humanoid robot dataset globally\u0026rdquo; Leju Robotics (乐聚智能) has focused on real-machine data for the post-training phase of model development, releasing datasets covering three core areas: whole-body motion control (including climbing windows and tables), dexterous hand manipulation, and basic upper-body movements The OpenLET dataset has achieved 3 million+ cross-platform downloads, and multiple embodied AI LLMs are citing OpenLET data to support model training Full-stack open source — from hardware to application:\nUnitree demonstrated a five-layer open source architecture: hardware base → data → toolchain → LLMs → applications Hardware: General robot platform with bottom-layer SDK, customizable down to individual motor SDKs (H2 humanoid robot: 31 degrees of freedom) Tools: Full development toolchain including teleoperation, visual data collection, and simulation training frameworks (GPU parallel training, reinforcement learning frameworks) Applications: UniStore — a \u0026ldquo;robot App Store\u0026rdquo; where developers can publish and share functional modules SoftPower (软通动力) advocates building a complete closed-loop from technology breakthrough to commercial deployment, based on open source root technology Ecosystem building — competitions, funds, and local working groups:\nCompetitions: Leju organizes 10+ competitions annually covering 600+ universities with 10,000+ participants. At ICRA 2026 in Austria, OpenAtom and Leju co-hosted the full-size humanoid robot challenge attracting 340+ teams globally, won by National University of Singapore Funds: A dedicated fund has been established to support community competitions and provide resources for entrepreneurial developers and students Local working groups: Working groups established in Ningbo, Hefei, Beijing, Zhengzhou, with 6-7 more planned. Each working group is led by a local academic heavyweight (Changjiang Scholar or National Outstanding Youth Science Fund recipient) Institutional infrastructure: The OpenAtom Foundation is described as playing the role of \u0026ldquo;institutional infrastructure\u0026rdquo; (制度基础设施) — from the OpenLET community to the M-Robots community, the foundation\u0026rsquo;s project incubation system is providing the governance framework for embodied AI open source development.\nSource: OpenAtom Foundation Journalism\n🏗️ Institutional Change 4. OpenAtom Foundation Prepares \u0026ldquo;AI Agent Open Source Working Group\u0026rdquo;\nAccording to Sogou WeChat search results, the OpenAtom Foundation is preparing to establish an \u0026ldquo;AI Agent Open Source Working Group\u0026rdquo; (智能体开源工作组). This follows the foundation\u0026rsquo;s recent focus on AI agent interconnection standards (GB/Z 185—2026, covered in the July 22 briefing) and the AIP open source reference implementation. The working group would serve as a dedicated governance body for AI agent-related open source projects within the foundation\u0026rsquo;s incubation system.\nSource: Sogou WeChat Search — OpenAtom Foundation\n5. OpenAtom Foundation Publishes Graduate Standards: \u0026ldquo;Lenient Entry, Strict Exit, Focus on Maturity\u0026rdquo;\nThe OpenAtom Foundation has released updated graduation standards for incubated projects, adopting a \u0026ldquo;lenient entry, strict exit, maturity-focused\u0026rdquo; (宽进严出、聚焦成熟度) approach. This represents a shift from quantity-based incubation to quality-based governance, aligning with the broader trend identified in the local policy analysis — moving from \u0026ldquo;building platforms\u0026rdquo; to \u0026ldquo;strengthening ecosystems.\u0026rdquo;\nSource: Sogou WeChat Search — OpenAtom Foundation\n🔍 WeChat Monitor OpenAtom Foundation Journalism (开放原子开源基金会新闻):\n2026-07-22: MIIT press conference on open source ecosystem (covered in July 24 briefing) 2026-07-21: \u0026ldquo;70+ Policies Behind: Local Open Source Enters \u0026lsquo;Value Realization Period\u0026rsquo;\u0026rdquo; — comprehensive analysis of local policy landscape 2026-07-20: \u0026ldquo;Embodied AI\u0026rsquo;s \u0026lsquo;Open Source Moment\u0026rsquo;\u0026rdquo; — OpenLET, Unitree, and Leju developments Sogou WeChat search trends:\nOpenAtom Foundation TOC discussions on \u0026ldquo;how to make foundations接地气 (down-to-earth) and how to do open source\u0026rdquo; OpenAtom Foundation\u0026rsquo;s new \u0026ldquo;AI Agent Open Source Working Group\u0026rdquo; in preparation OpenAtom Foundation\u0026rsquo;s new graduate standards for incubated projects 🔍 Commentary The \u0026ldquo;Value Realization Period\u0026rdquo; — Institutional Economics of Chinese Local Open Source Policy\nThe OpenAtom Foundation\u0026rsquo;s July 21 analysis of local open source policies is arguably the most important institutional document published this week. It provides a framework for understanding the phase transition Chinese open source is undergoing at the subnational level.\n1. From \u0026ldquo;Policy Quantity\u0026rdquo; to \u0026ldquo;Policy Quality\u0026rdquo; — The Institutional Maturation Signal\nThe 70+ local policies containing open source content represent a critical mass. In institutional economics terms, this is the point at which a policy innovation moves from \u0026ldquo;experimental\u0026rdquo; to \u0026ldquo;institutionalized.\u0026rdquo; The analysis\u0026rsquo;s key insight is that the real question is no longer \u0026ldquo;how many policies?\u0026rdquo; but \u0026ldquo;do they actually work?\u0026rdquo;\nThe four shifts identified — embedded into industrial systems, full-element openness, dual online+offline centers, and industrial scenario empowerment — represent a maturation of institutional design. Early open source policies were copies of each other (incubation platforms, developer conferences, industry funds). The new generation is differentiated by local industrial context. This is a classic institutional economics pattern: imitation gives way to adaptation as the costs of homogeneity become apparent.\n2. Zhejiang\u0026rsquo;s Policy as Institutional Innovation\nThe Zhejiang 7-department plan is significant for several reasons:\nFirst, it represents inter-agency coordination across seven departments — a level of institutional integration that most local open source policies have not achieved. This is not a single-ministry directive but a whole-of-government approach.\nSecond, the policy instruments selected — SBOM requirements, computing vouchers, data vouchers, government procurement, and contribution-based talent evaluation — are institutionally sophisticated. They move beyond the standard toolkit of \u0026ldquo;funding and events\u0026rdquo; to create actual incentive structures and market mechanisms.\nThird, the 2027 targets (50+ projects, 150+ enterprises, 100+ benchmarks) are measurable and time-bound — a governance approach that creates accountability and enables evaluation.\n3. The Three Contradictions — Supply-Side vs. Demand-Side in Open Source Governance\nThe analysis identifies three structural contradictions that are remarkably honest for a foundation-affiliated publication:\nSupply convergence vs. demand differentiation: The \u0026ldquo;one-size-fits-all\u0026rdquo; problem in local open source policy is a classic public goods problem — governments tend to provide standardized solutions, but industries need customized ones Construction vs. operations: The \u0026ldquo;build but cannot run\u0026rdquo; (建得起、转不动) problem is a governance capacity issue — building infrastructure is easier than building the institutional capacity to sustain it Talent evaluation vs. open source contribution: This is perhaps the most fundamental institutional barrier. The current Chinese talent evaluation system (职称评定, 人才计划) is based on traditional academic and professional metrics that do not recognize distributed, community-based contributions. Until this changes, the incentive structure for sustained open source participation will remain misaligned 4. Embodied AI and the Institutional Infrastructure Thesis\nThe embodied AI article provides a concrete case study of the OpenAtom Foundation\u0026rsquo;s role as \u0026ldquo;institutional infrastructure.\u0026rdquo; The OpenLET community, Unitree\u0026rsquo;s open source data, Leju\u0026rsquo;s competition-and-fund model, and the foundation\u0026rsquo;s incubation framework represent a multi-layered institutional ecosystem:\nTechnical layer: Open datasets, toolchains, and hardware interfaces Community layer: Competitions, working groups, and developer networks Governance layer: The OpenAtom Foundation\u0026rsquo;s project lifecycle management Capital layer: Dedicated funds for open source entrepreneurship This layered structure is precisely what the \u0026ldquo;value realization\u0026rdquo; thesis predicts — open source ceases to be a single-dimensional activity and becomes a multi-dimensional institutional arrangement that coordinates technical development, community governance, talent cultivation, and capital allocation.\n5. The Week Ahead\nThe convergence of three developments — the local policy analysis, the Zhejiang plan, and the embodied AI open source momentum — suggests that Chinese open source institutional development is accelerating at the subnational level. While the national-level story this week has been dominated by the US-China AI sanctions confrontation (July 22-24 briefings), the local-level story is one of pragmatic institutional construction. The sanctions narrative may dominate headlines, but the policy infrastructure being built at the provincial level may have a more durable impact on the long-term trajectory of Chinese open source.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-27/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-27\"\u003eChina Open Source Daily — 2026-07-27\u003c/h2\u003e\n\u003ch3 id=\"-policy--regulation\"\u003e🏛️ Policy \u0026amp; Regulation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. 70+ Local Open Source Policies: China\u0026rsquo;s Local Open Source Enters \u0026ldquo;Value Realization Period\u0026rdquo;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe OpenAtom Foundation published a comprehensive analysis on July 21 documenting the evolution of China\u0026rsquo;s local open source policy landscape. According to the analysis, the number of local policies explicitly containing \u0026ldquo;open source\u0026rdquo; content has exceeded \u003cstrong\u003e70 nationwide\u003c/strong\u003e, marking a critical inflection point for the institutionalization of open source at the subnational level.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-27"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-24 🏛️ Policy \u0026amp; Regulation 1. MIIT Releases First Official Quantification of China\u0026rsquo;s Open Source Ecosystem — 1.35 Billion OpenHarmony Devices, 10 Billion Global AI Model Downloads\nAt the State Council Information Office press conference on July 20, MIIT spokesperson and Director of the Operation Monitoring and Coordination Bureau Tao Qing provided the first comprehensive official quantification of China\u0026rsquo;s open source ecosystem during the 2026 H1工业和信息化发展情况 (Industrial and Information Technology Development) briefing.\nKey data points released by MIIT:\nMetric Value OpenHarmony ecosystem devices 1.35 billion+ OpenHarmony-based industry distributions 100+ (including \u0026ldquo;Electric IoT OS\u0026rdquo; and \u0026ldquo;Instrument IoT OS\u0026rdquo;) National foundation projects in incubation 50+ National AI open source community users 11 million+ Models hosted on national AI open source platform 70,000+ Global downloads of Chinese AI open-source models 10 billion+ Software revenue (Jan-May 2026) ¥6.2 trillion, +10.3% YoY Enterprise digital R\u0026amp;D tool adoption rate 86.3% Key process CNC rate 69.5% Tao Qing specifically addressed the question of \u0026ldquo;how to cultivate the open source ecosystem\u0026rdquo; in response to a journalist\u0026rsquo;s query, making four key points:\nBasic software accelerating: OpenHarmony now covers smartphones, computers, cars, and home appliances — a full-stack coverage from IoT to desktop Industrial software widely deployed: Coverage across key industries, reaching 86.3% digital R\u0026amp;D tool adoption AI fully empowering: Software enterprises adopting code LLMs and intelligent programming tools at scale Open source ecosystem accelerating: The national foundation has incubated 50+ projects; the national AI open source community now hosts 70,000+ models with 11M+ users Open source as explicit policy priority: Tao Qing stated that MIIT will:\nAccelerate the introduction of an \u0026ldquo;AI + Software\u0026rdquo; action plan (人工智能+软件行动方案) Promote intelligent transformation of software development, smart upgrade of software products/services, and cultivation of AI agent software as a new business category Enhance open source infrastructure in project incubation, supply chain security governance, emerging scenario exploration, and overseas market landing Accelerate the development of the national AI open source community into a \u0026ldquo;source of innovation for emerging technologies and an important scenario for software-hardware product deployment\u0026rdquo; Institutional significance: This is the first time China\u0026rsquo;s government has released comprehensive, quantified data on its open source ecosystem through a formal State Council press conference. The framing is significant: open source is presented as an integrated component of industrial policy — not merely a development methodology but a strategic infrastructure for software, AI, and hardware ecosystem development. The explicit mention of \u0026ldquo;overseas market landing\u0026rdquo; as a policy priority for open source infrastructure signals that Chinese open source is now a vehicle for global technology deployment, not just domestic innovation.\nSources: China News Service, AIBase (English summary), State Council Information Office\n2. \u0026ldquo;AI + Software\u0026rdquo; Action Plan — A New Industrial Policy Instrument for Open Source\nBeyond the data, the MIIT press conference revealed a significant new policy instrument: the \u0026ldquo;AI + Software\u0026rdquo; action plan (人工智能+软件行动方案). Unlike previous \u0026ldquo;Software + AI\u0026rdquo; framings which treated AI as a bolt-on feature, the new formulation positions AI as the primary driver of software industry transformation.\nThree pillars of the action plan:\nSoftware development intelligent transformation: Code LLMs, AI programming tools, and automated testing/operations Software product and service intelligent upgrade: Smart features, natural language interaction, AI-generated content AI agent software as new business category: The \u0026ldquo;AaaS\u0026rdquo; (Agent as a Service) model — consistent with the ¥3.3 trillion AI agent market projection cited in the Xinhua/OpenAtom analysis from July 22 The action plan is explicitly linked to open source infrastructure: Tao Qing stated that the national AI open source community will be the \u0026ldquo;source of innovation\u0026rdquo; for the plan. This creates a direct institutional link between open source governance and industrial policy implementation — a distinctive Chinese approach that differs from the Western model where open source and industrial policy operate in separate domains.\nSource: China News Service\n⚖️ Legal \u0026amp; Licensing 3. White House Specifically Accuses Moonshot AI of Distilling Anthropic\u0026rsquo;s Fable 5 for Kimi K3 — Treasury Sanctions Threat Intensifies\nOn July 22-23, the White House escalated its confrontation with Chinese open-source AI by making specific allegations against Moonshot AI\u0026rsquo;s Kimi K3 model:\nMichael Kratsios, Director of the White House Office of Science and Technology Policy (OSTP), publicly accused Moonshot AI of using covert distillation of Anthropic\u0026rsquo;s Fable 5 to train Kimi K3 — a 2.8-trillion-parameter open-weight model released on July 16. Kratsios also alleged that Moonshot accessed Nvidia\u0026rsquo;s Blackwell GPUs despite export controls.\nUS Treasury Secretary Scott Bessent doubled down on his sanctions threat (first reported July 21), stating that sanctions remain on the table and that the administration is \u0026ldquo;carefully examining\u0026rdquo; Chinese open-source AI models for IP theft.\nKey details of the allegations:\nKratsios stated on X that the US government \u0026ldquo;holds information confirming Moonshot AI distilled capabilities from Anthropic\u0026rsquo;s Fable AI during development of Kimi K3\u0026rdquo; The accusation draws a line between open-weight releases (which the administration says it supports) and covert extraction of proprietary model outputs (which it calls theft) Moonshot\u0026rsquo;s K3 is scheduled for full open-weight release on July 27 Anthropic\u0026rsquo;s Fable 5 was re-released on July 1 after being taken offline due to US export controls, giving a narrow window for distillation claims Some AI researchers have questioned the claims, noting that Kimi K3\u0026rsquo;s architecture differs significantly from Fable 5 Institutional significance: The framing by Bessent is strategically important — he explicitly distinguishes between \u0026ldquo;legitimate open-source AI\u0026rdquo; and \u0026ldquo;stolen IP released as open source.\u0026rdquo; This is an attempt to create a legal distinction that would allow the US to sanction specific open-weight models without appearing to oppose open source in general. If this distinction becomes legally operationalized, it would create a new category of \u0026ldquo;tainted open source\u0026rdquo; — models that are open in form but sanctioned in substance. The burden of proof shifts to model creators to demonstrate that their training data and methods are clean, an extremely difficult standard for open-weight releases.\nSources: TechCrunch, CNBC, The Register, PCMag, ExplainX, ZeroHedge\n🏗️ Institutional Change 4. Jensen Huang Defends Chinese Open-Source AI — A Pivotal Intervention from the Industry\u0026rsquo;s Most Influential CEO\nIn an exclusive interview with Axios on July 22, Nvidia CEO Jensen Huang made a forceful defense of Chinese open-source AI models, directly challenging the US administration\u0026rsquo;s escalating sanctions campaign:\n\u0026ldquo;These Chinese models are excellent. Open-source models that are excellent should be used.\u0026rdquo;\nKey arguments from Huang\u0026rsquo;s interview:\nOpenness = security: Huang argued that open models are more secure, not less, because outside researchers can inspect them, expose weaknesses, and build defenses Monoculture risk: \u0026ldquo;If everything just becomes one single model, one single point of attack, one single source of failure, I think the world is much less safe\u0026rdquo; — a direct argument against the proprietary model approach Market reality: Chinese open-source models are already widely used by enterprises globally, and bans would disrupt existing deployments Business logic: Open-source AI expands the overall market for AI infrastructure, including Nvidia\u0026rsquo;s chips and data centers — a self-interested but structurally significant argument Institutional significance: Huang\u0026rsquo;s intervention is remarkable for several reasons:\nNvidia\u0026rsquo;s strategic position: Nvidia is the primary beneficiary of the AI infrastructure buildout on both sides of the Pacific. Huang has more to lose from AI fragmentation than almost any other industry figure The timing: His statement comes just as the White House and Treasury Department are escalating sanctions — a direct counter-signal from the industry\u0026rsquo;s most powerful voice The framing: Huang doesn\u0026rsquo;t argue that Chinese models are not competitive with US models (which would be defensive). He argues they are excellent and should be embraced — an offensive reframing that positions openness as a strength, not a vulnerability The global audience: The interview was published in Axios, Fortune, Yahoo Finance, and MoneyControl — reaching both US policymakers and global investors This is the most significant industry voice to push back against the US sanctions campaign against Chinese open-source AI, and it signals that the \u0026ldquo;containment faction\u0026rdquo; and \u0026ldquo;engagement faction\u0026rdquo; in US AI policy now have a major corporate champion.\nSources: Axios, Fortune, Yahoo Finance, MoneyControl, TimesNow\n5. Arcee (US Open Source AI Lab): \u0026ldquo;Chinese Models Are Not Inherently Dangerous\u0026rdquo;\nOn July 22, Arcee — a US-based open-source AI company — published a notable counterpoint to the Washington panic. CTO Lucas Atkins stated clearly:\n\u0026ldquo;Chinese models are no more dangerous than any other open source software a company may use.\u0026rdquo;\nKey arguments:\nDownload-and-run security: Once a model is downloaded and run in an enterprise\u0026rsquo;s own environment, the original developer has no access to it — the model is just weights, not a remote service Open source != espionage: Open-weight models cannot \u0026ldquo;phone home\u0026rdquo; or exfiltrate data unless the user explicitly configures them to do so Inspection advantage: Open-source models can be audited, inspected, and modified by the user — unlike proprietary models where the internal workings are opaque Practical benefit: Arcee itself uses Chinese open-source models as base models for fine-tuning, finding them technically competitive with US alternatives Institutional significance: Arcee\u0026rsquo;s statement is notable because it comes from a US company that competes in the same market as Chinese AI labs. Unlike Jensen Huang (who has commercial interests in chip sales), Arcee\u0026rsquo;s argument is purely technical and practical. The company is effectively saying: \u0026ldquo;The open-source model is working as designed — we inspect, we use, we build on top. The security concerns are not technical but political.\u0026rdquo;\nThis represents a growing divide between US policy makers (who view Chinese open-source AI as a national security threat) and US technology practitioners (who view it as a useful resource). The Arcee position, combined with Jensen Huang\u0026rsquo;s, suggests that the US open-source AI community is actively resisting the securitization of open-weight model distribution.\nSources: TechCrunch, BitcoinWorld, HotON.ai\n🔍 WeChat Monitor 6. OpenAtom Foundation Journalism — MIIT Press Conference Coverage\nThe OpenAtom Foundation\u0026rsquo;s official website prominently features the MIIT press conference results on its homepage, framing the 1.35 billion OpenHarmony devices and 10 billion AI model downloads as validation of the foundation\u0026rsquo;s incubation model. The foundation\u0026rsquo;s journalism page has been updated with the July 20 press conference as the lead story, signaling that the foundation views this official quantification as a significant legitimization event.\nSource: OpenAtom Foundation\n🔍 Commentary The Week of Official Quantification: Three Signals for Chinese Open Source\n1. The MIIT Press Conference Represents a Governance Phase Transition\nThe July 20 press conference is the first time China\u0026rsquo;s government has released quantified open source data through a formal State Council briefing. This is not a technical report from a standards body — it is a ministerial-level policy statement that treats open source as a formal category of industrial output, measured alongside software revenue (¥6.2 trillion) and digitalization rates (86.3%).\nThe institutional significance cannot be overstated: open source has been officially recognized as a measurable component of China\u0026rsquo;s technology economy. The metrics — 1.35 billion devices, 10 billion downloads, 50+ incubated projects, 11 million platform users — create a baseline against which future policy will be evaluated. This is the same governance logic that underlies the OpenAtom Deep Report (covered July 22): what gets measured gets managed.\n2. The \u0026ldquo;AI + Software\u0026rdquo; Plan Creates a New Institutional Link\nThe \u0026ldquo;AI + Software\u0026rdquo; action plan, with its explicit linkage to the national AI open source community, represents a Chinese innovation in industrial policy architecture. In the Western model, open source foundations and industrial policy operate in separate spheres — foundations govern code, governments set market conditions. In the Chinese model, the national AI open source community is explicitly designated as the \u0026ldquo;source of innovation\u0026rdquo; for a government industrial policy action plan.\nThis is neither top-down control (the community has 11 million users, not a government-appointed board) nor bottom-up autonomy (the policy explicitly directs funding and attention to the platform). It is a hybrid governance model that the Chinese system is uniquely equipped to implement: a government-funded, community-operated, industry-oriented open source platform that serves as a policy instrument without being a government agency.\n3. The Convergence of the Sanctions Narrative and the Quantification Narrative\nThis week presents a remarkable juxtaposition: on the same days that the US Treasury threatens sanctions on Chinese open-source AI models (July 21-22) and the White House specifically accuses Moonshot AI of theft (July 22-23), the Chinese government publishes its first official quantification of the open source ecosystem it has built (July 20).\nThe MIIT\u0026rsquo;s 10 billion global download figure for Chinese AI open-source models is a direct counterpoint to the US narrative. The data says: these models are being used, everywhere, by everyone — they are not theft, they are adoption. The 29 nations that signed WAICO (covered July 23) are the diplomatic manifestation of the same argument: Chinese open-source AI is a global phenomenon, not a covert operation.\nThe institutional insight: China is now fighting the narrative war with data, not just rhetoric. The State Council press conference provided the data; WAICO provided the diplomatic vehicle; the Mulan License donation (covered July 23) provided the legal infrastructure. The three developments together form a coherent institutional response to the US sanctions campaign: quantify, legitimize, govern.\n4. The Industry Counter-Narrative Takes Shape\nJensen Huang and Arcee represent a significant development: the US technology industry is beginning to push back against the sanctions narrative. Two arguments are emerging:\nSecurity through openness (Huang): Open models are more secure because they can be inspected. Banning them makes the ecosystem less safe. Technical normalcy (Arcee): Chinese open-source models are just software — no more dangerous than any other open-source dependency. These arguments are structurally important because they align the interests of the US open-source community with the Chinese open-source ecosystem. If the US sanctions campaign succeeds, it will harm not only Chinese AI labs but also US companies that build on Chinese open-source models. The industry pushback suggests that the \u0026ldquo;bilateral enclosure of the open-source AI commons\u0026rdquo; (identified in the July 22 commentary) has a powerful constituency on the US side that will resist it.\n📊 Trends 1. The Official Quantification of Chinese Open Source The MIIT press conference establishes a new governance baseline: Chinese open source is now measured, reported, and managed as a formal component of technology industrial policy. Expect annual or semi-annual updates to these metrics, creating a data series that will inform both domestic policy and international positioning.\n2. The \u0026ldquo;AI + Software\u0026rdquo; Action Plan as a Policy Instrument This is the first time a major Chinese government action plan explicitly designates an open source community as its institutional foundation. If successful, this model could be replicated for other technology domains (AI + manufacturing, AI + healthcare, etc.), creating a new category of policy-aligned open source infrastructure.\n3. The Industry-Policy Divide in US AI Governance The divergence between Jensen Huang/Arcee (industry) and the White House/Treasury (policy) on Chinese open-source AI represents a growing institutional fracture. The US has no equivalent of China\u0026rsquo;s centralized quantification or coordinated policy response — instead, it has competing factions with incompatible interests. This institutional fragmentation is a structural advantage for China\u0026rsquo;s coordinated approach.\n4. The \u0026ldquo;Tainted Open Source\u0026rdquo; Legal Precedent If the US Treasury establishes a sanctions regime that distinguishes between \u0026ldquo;legitimate\u0026rdquo; and \u0026ldquo;tainted\u0026rdquo; open-source model weights, it will create a legal category with no precedent in open-source governance. The open-source community has never had to certify that its training data was not \u0026ldquo;stolen\u0026rdquo; — the concept is foreign to the legal framework of open-source licensing. A sanctions regime that requires such certification would fundamentally alter the governance of open-weight AI distribution.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-24/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-24\"\u003eChina Open Source Daily — 2026-07-24\u003c/h2\u003e\n\u003ch3 id=\"-policy--regulation\"\u003e🏛️ Policy \u0026amp; Regulation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. MIIT Releases First Official Quantification of China\u0026rsquo;s Open Source Ecosystem — 1.35 Billion OpenHarmony Devices, 10 Billion Global AI Model Downloads\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAt the \u003cstrong\u003eState Council Information Office press conference on July 20\u003c/strong\u003e, MIIT spokesperson and Director of the Operation Monitoring and Coordination Bureau \u003cstrong\u003eTao Qing\u003c/strong\u003e provided the first comprehensive official quantification of China\u0026rsquo;s open source ecosystem during the 2026 H1工业和信息化发展情况 (Industrial and Information Technology Development) briefing.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-24"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-23 🏛️ Policy \u0026amp; Regulation 1. China Launches WAICO — 29-Nation AI Governance Bloc Backing Open-Source AI\nOn July 17, Chinese President Xi Jinping announced the creation of the World Artificial Intelligence Cooperation Organization (WAICO) at the opening ceremony of the 2026 World AI Conference (WAIC) in Shanghai, with representatives of 29 nations signing the founding agreement. UN Secretary-General António Guterres attended the ceremony.\nWAICO represents China\u0026rsquo;s most ambitious institutional play in global AI governance — a treaty-based intergovernmental organization headquartered in Shanghai, explicitly designed to offer an alternative governance framework to the US-led approach. Key announcements:\nOpen-source AI as a global public good: Xi\u0026rsquo;s keynote framed open-source AI as a tool for bridging the Global North-South AI divide, directly contrasting with US export controls and sanctions 5,000 AI training slots: China pledged 5,000 AI training opportunities for developing nations over the next five years Founding members: Drawn primarily from the Global South — Pakistan, Indonesia, Thailand, Kazakhstan, Brazil, Saudi Arabia, UAE, and others. India notably declined to join, and no Western democratic nations were signatories Institutional structure: WAICO will have a secretariat in Shanghai, with working groups on standards, safety, capacity building, and open-source cooperation The Diplomat\u0026rsquo;s analysis (July 21) notes that WAICO\u0026rsquo;s primary function is \u0026ldquo;the collective normalization of China\u0026rsquo;s approaches to AI governance\u0026rdquo; — participating countries will then carry China\u0026rsquo;s preferred norms into other international forums. The organization positions China as the champion of open-source AI as development infrastructure, in direct competition with the US model of proprietary AI leadership.\nSignificance: WAICO is the most significant institutional development in Chinese open-source AI governance since the Global AI Governance Initiative (2023). By creating a treaty-based organization with 29 nations before the UN\u0026rsquo;s own AI governance process has matured, China has established an institutional vehicle that can: (a) set technical standards for open-source AI interoperability, (b) define acceptable use norms for open-weight models, and (c) channel Chinese open-source AI models to Global South markets as an alternative to US-controlled platforms.\nSources: People\u0026rsquo;s Daily, Eastern Herald, The Diplomat, IBTimes, Medianama, Model Diplomat\n2. China Weighs Export Controls on Its Own Frontier AI Models — Including Open-Weight Releases\nOn July 7, 2026, Reuters exclusively reported that China\u0026rsquo;s Ministry of Commerce had convened meetings with Alibaba, ByteDance, and Z.ai (Zhipu) to discuss potentially restricting overseas access to China\u0026rsquo;s most advanced AI models — including open-weight releases and models not yet released.\nKey details from the report:\nThe discussions are at an early, exploratory stage Proposed restrictions could cover model weights, API access, and technical documentation The framework would be China\u0026rsquo;s equivalent of the US export controls on AI models Chinese tech companies reportedly expressed concerns about losing global developer mindshare if they comply Institutional significance: This development creates a fascinating symmetry with the US Treasury sanctions threat (covered in the July 22 briefing). Both the US and China are now exploring restrictions on open-weight AI model distribution — but for diametrically opposed reasons. The US wants to prevent Chinese models from competing with American AI; China wants to prevent its models from being used in ways that undermine its own technological advantage. The result is a bilateral tightening of the open-source AI commons from both directions.\nSources: TIME, ExplainX, Yahoo News, ZGLG\n⚖️ Legal \u0026amp; Licensing 3. Mulan License Family Formally Donated to OpenAtom Foundation — China Open Source Enters \u0026ldquo;Rule Export\u0026rdquo; Era\nAt the 2026 OpenAtom Open Source Ecosystem Conference (June 25, Beijing), the Mulan License Family — comprising 6 licenses for different use scenarios — was formally donated to the OpenAtom Foundation. This marks a critical institutional transition for China\u0026rsquo;s indigenous open-source licensing framework.\nBackground on the Mulan License Family:\nMulanPSL-2.0 (Mulan Permissive Software License v2) — OSI-approved in 2020, the first internationally recognized Chinese open-source license. More permissive than Apache 2.0, with simplified Chinese-language legal text MulanPubL-2.0 (Mulan Public License v2) — A copyleft license comparable to GPL, used by OceanBase (Ant Group\u0026rsquo;s distributed database) and other major Chinese open-source projects 4 additional licenses covering different scenarios (academic, hardware, data, etc.) The licenses were developed by: China Electronics Standardization Institute (CESI), Peking University, National University of Defense Technology, and Shanghai Jiao Tong University.\nPost-donation governance model: A joint working group established by the OpenAtom Foundation and CESI will jointly manage:\nLicense version maintenance and updates OSI certification stewardship Compliance guidance and dispute resolution Adoption promotion across Chinese open-source projects Institutional significance: This is arguably the most important institutional development in Chinese open-source licensing since the OSI\u0026rsquo;s approval of MulanPSL-2.0 in 2020. By moving the Mulan License Family from academic/standards-body governance to foundation governance under OpenAtom, China achieves three things:\nCredible neutral governance: The licenses are no longer controlled by a government standards body (CESI) but by a foundation that follows open governance norms Licensing sovereignty: Chinese open-source projects can now use a full license family (from permissive to strong copyleft) that is governed domestically, reducing dependence on US-centric licenses (MIT, Apache, GPL) that carry US legal jurisdiction Rule export capacity: The foundation can now promote the Mulan licenses internationally — analogous to how Creative Commons licenses became a global standard through institutional governance rather than government mandate The InfoQ headline captured this: \u0026ldquo;Mulan License Family \u0026lsquo;Six-Piece Set\u0026rsquo; Formally Donated to Open Source Foundation — China Open Source Enters the Era of Rule Export.\u0026rdquo;\nSources: InfoQ, Baidu Baike - Mulan License Family, ithome, Baidu Baike - Mulan Public License\n4. China\u0026rsquo;s Anthropomorphic AI Interaction Rules Take Effect July 15 — Regulating AI Companions and Emotional Chatbots\nOn July 15, 2026, China\u0026rsquo;s Interim Measures for the Administration of Anthropomorphic AI Interaction Services officially took effect. Issued jointly by the Cyberspace Administration of China (CAC), NDRC, MIIT, Ministry of Public Security, and SAMR on April 10, 2026, the regulations represent the world\u0026rsquo;s first comprehensive regulatory framework specifically targeting AI systems that simulate human-like emotional interaction.\nKey requirements:\nBan on romantic/emotional dependency design: Services cannot be designed to create emotional dependence, particularly for minors Mandatory pre-launch review: Anthropomorphic AI services must undergo content safety review before release Real-time content moderation: Providers must implement filtering for prohibited content, including politically sensitive topics Minor protection: Strict age verification and restricted interaction patterns for underage users Transparency labeling: AI systems must clearly identify themselves as AI, not human The regulations sparked widespread public reaction — TechXplore reported that Chinese users of AI companion bots held \u0026ldquo;heart-rending farewells\u0026rdquo; to their virtual buddies as the rules took effect, particularly popular AI companion apps that offered romantic or family-like interaction.\nOpen-source implications: The rules apply to both proprietary and open-source AI companion systems deployed in China. For open-source AI projects that include anthropomorphic interaction capabilities, the regulations create compliance obligations around content filtering, age verification, and transparency — requirements that may be difficult to implement in fully open-source, decentralized models.\nSources: TechXplore, Hogan Lovells, AI Governance, Geopolitechs, Licentium, Times of AI\n🏗️ Institutional Change 5. The Diplomat: China\u0026rsquo;s New AI Governance Organization Seeks to Formalize Global Influence\nThe Diplomat published a strategic analysis (July 21) examining WAICO\u0026rsquo;s institutional architecture and its implications for global AI governance. Key analytical points:\nTiming advantage: China launched WAICO (July 2026) before the UN\u0026rsquo;s AI governance process matures, establishing its preferred norms as a baseline for 29 nations Open-source as diplomatic currency: By offering open-source AI models, training, and infrastructure to Global South countries, China provides tangible benefits that Western governance frameworks (which focus on safety and ethics) do not The normalization mechanism: WAICO\u0026rsquo;s real function is not to set hard rules but to \u0026ldquo;normalize\u0026rdquo; China\u0026rsquo;s approach to AI governance across participating countries, who then carry those norms into other forums The India question: India\u0026rsquo;s absence from WAICO is significant — the two Asian AI powers are now competing for Global South AI allegiance through separate governance vehicles Source: The Diplomat\n🔍 WeChat Monitor 6. OpenAtom Open Source Ecosystem Conference — Mulan License Donation \u0026amp; Project Donations\nThe OpenAtom Foundation\u0026rsquo;s official journalism page (via openatom.org) reports on the June 25 conference, where besides the Mulan License donation, several projects completed formal donation agreements to the foundation, including the Electric IoT OS (电鸿) and other industry-specific open-source platforms. These project donations follow the standard OpenAtom incubation pipeline: code contribution → technical review → legal due diligence → foundation governance transition.\nSource: OpenAtom Foundation Journalism\n🔍 Commentary The Week China Reshaped Global Open-Source Governance\nThe period between July 7 and July 23, 2026, may be remembered as the inflection point when Chinese open-source governance moved from a domestic industrial policy to a global institutional strategy. Four developments, each significant individually, collectively represent a phase transition:\n1. The Two-Sided Enclosure of the Open-Source AI Commons\nThe US Treasury Secretary threatens sanctions on Chinese open-source models (July 21). China\u0026rsquo;s Ministry of Commerce explores export controls on its own models (July 7). Both the world\u0026rsquo;s largest AI powers are now considering restricting the distribution of open-weight AI models — but from opposite directions. The result is a bilateral enclosure of what was, until recently, considered a global commons. The open-source AI community, which has built its governance model on the premise of frictionless global distribution, now faces a world where both the US and China are erecting barriers.\n2. WAICO as Institutional Entrepreneurship\nChina\u0026rsquo;s launch of WAICO with 29 nations is a textbook case of institutional entrepreneurship — the creation of new institutional arrangements that reshape the rules of the game. By making open-source AI the centerpiece of its diplomatic offering to the Global South, China has achieved something remarkable: the country most accused of IP theft is now the leading state-level advocate for open-source AI as a global public good. This is the same playbook used by Creative Commons and the Open Source Initiative — but deployed at the intergovernmental level. The key insight: you don\u0026rsquo;t need to win the argument in Geneva if you\u0026rsquo;ve already won it in Shanghai.\n3. The Mulan License Donation: From Domestic Standard to Foundation Governance\nThe donation of the Mulan License Family to OpenAtom represents the maturation of China\u0026rsquo;s open-source licensing infrastructure. The significance is not the licenses themselves — they are functionally similar to Apache 2.0 and GPL — but the institutional architecture around them. By anchoring the Mulan licenses in a foundation (OpenAtom) rather than a government standards body (CESI), China creates a governance structure that is:\nCredible to international developers: Foundation governance, not government control Sovereign in legal terms: Chinese law jurisdiction, not US law Scalable: The foundation can maintain, update, and promote the licenses globally This is China\u0026rsquo;s answer to the question: \u0026ldquo;How do you participate in global open source while maintaining regulatory autonomy?\u0026rdquo; The answer: build your own license family, governed by your own foundation, but interoperable with the global ecosystem.\n4. The Emerging Pattern: A Two-Track Global Open-Source AI System\nTaken together, these developments point toward a two-track global open-source AI system:\nTrack 1 (China-led): WAICO member states, Mulan-licensed models, OpenAtom-governed projects, AI training and infrastructure provided by Chinese companies Track 2 (US-led): US and allied nations, traditional licenses (Apache, MIT, GPL), Linux Foundation-governed projects, models distributed through US platforms The two tracks are not fully separate — there is significant overlap in developer communities, code reuse, and technical standards — but the governance infrastructure is diverging. The question for the next 12 months is whether this divergence creates productive competition (two models of open-source AI governance competing for legitimacy) or destructive fragmentation (incompatible legal and technical standards that Balkanize the global AI ecosystem).\n5. The Paradox of China\u0026rsquo;s Anthropomorphic AI Regulation\nChina\u0026rsquo;s new AI companion rules present a paradox for open-source advocates. On one hand, the regulations represent the kind of proactive governance that the open-source community often calls for — clear rules, transparency requirements, child protection. On the other hand, the compliance burden falls disproportionately on open-source projects, which lack the legal and engineering resources of large companies. This tension — between the desire for regulatory clarity and the reality of regulatory burden on decentralized communities — will be a defining challenge for Chinese open-source AI governance in the coming years.\n📊 Trends 1. The Institutionalization of the Open-Source AI Geopolitics WAICO, Mulan licensing, and export controls on both sides of the Pacific constitute a structural shift from open-source AI as a technical community to open-source AI as a domain of interstate competition. The governance of open-source AI is no longer primarily a matter of community norms — it is now a matter of treaty obligations, sanctions regimes, and export control law.\n2. The \u0026ldquo;Rule Export\u0026rdquo; Phase of Chinese Open Source The Mulan License donation marks the transition from \u0026ldquo;importing\u0026rdquo; open-source governance norms (adopting Apache, MIT, GPL) to \u0026ldquo;exporting\u0026rdquo; Chinese governance norms through foundation-based licensing infrastructure. This is the institutional parallel of the technological transition from \u0026ldquo;catching up\u0026rdquo; to \u0026ldquo;leading\u0026rdquo; in AI.\n3. The Global South as the Battleground WAICO\u0026rsquo;s 29 founding nations are primarily from the Global South. The competition for AI governance is now being fought not in the capitals of the US and China, but in the policy preferences of developing nations. Open-source AI is the primary weapon in this competition — the side that provides the most useful, accessible, and trusted open-source AI infrastructure wins the allegiance of the Global South\u0026rsquo;s AI ecosystem.\n4. The Convergence of US and China on Export Controls Perhaps the most ironic development: the US and China are converging on the same policy instrument — export controls on AI model weights — for diametrically opposed reasons. This convergence suggests that the open-source AI commons, which thrived on the assumption of frictionless global distribution, is entering an era of managed access from both sides of the geopolitical divide.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-23/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-23\"\u003eChina Open Source Daily — 2026-07-23\u003c/h2\u003e\n\u003ch3 id=\"-policy--regulation\"\u003e🏛️ Policy \u0026amp; Regulation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. China Launches WAICO — 29-Nation AI Governance Bloc Backing Open-Source AI\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 17, Chinese President Xi Jinping announced the creation of the \u003cstrong\u003eWorld Artificial Intelligence Cooperation Organization (WAICO)\u003c/strong\u003e at the opening ceremony of the 2026 World AI Conference (WAIC) in Shanghai, with representatives of 29 nations signing the founding agreement. UN Secretary-General António Guterres attended the ceremony.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-23"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-22 🏛️ Policy \u0026amp; Regulation 1. China Open Source Development Deep Report (2025) Released — AI Enters \u0026ldquo;Ecosystem Decisive\u0026rdquo; Era\nThe OpenAtom Foundation published the China Open Source Development Deep Report (2025), a comprehensive study based on data from GitHub, Gitee, and AtomGit. The report\u0026rsquo;s core thesis: global AI competition is shifting from \u0026ldquo;model比拼\u0026rdquo; to \u0026ldquo;ecosystem decisive\u0026rdquo;.\nKey data points:\nChina\u0026rsquo;s open source contributions: 77.98 million contributions, up 54.92% year-over-year, accounting for 12.38% of global share (up from 10.10% in 2023), ranking third globally behind the US and India. China\u0026rsquo;s active open source projects: 4.25 million, up 40.58% YoY. Global AI open source projects: 790,200 active projects, up 70.59% YoY; active AI developers up 62.85%. China\u0026rsquo;s AI open source projects: 95,200, accounting for 12.06% of global total, ranking third. Chinese universities dominate the global Top 30 AI open source developer rankings: 15 out of 30 slots, with Tsinghua, Zhejiang University, Shanghai Jiao Tong, and Peking University in the top five globally. The report makes a critical institutional observation: OSPOs are evolving from \u0026ldquo;compliance support departments\u0026rdquo; to organization-level infrastructure connecting enterprise R\u0026amp;D with external innovation ecosystems. In the AI era, OSPOs must now manage models, datasets, evaluations, inference frameworks, and application assets — not just software components and licenses.\nThe report also notes that open source infrastructure is shifting from \u0026ldquo;code repositories\u0026rdquo; to \u0026ldquo;intelligent platforms\u0026rdquo; — AtomGit is evolving into a \u0026ldquo;code + model + compute\u0026rdquo; integrated platform, representing a transition from \u0026ldquo;hosting projects\u0026rdquo; to \u0026ldquo;organizing innovation.\u0026rdquo;\nSource: OpenAtom Foundation Journalism\n2. AIP Open Source Project Launched Based on National AI Agent Interconnection Standard (GB/Z 185—2026)\nOn July 7, the AIP (Agent Interconnection Protocol) open source reference implementation was formally released on AtomGit, based on the newly published national standard GB/Z 185—2026 Artificial Intelligence — Agent Interconnection. The standard, comprising 7 guidance documents, addresses the lack of unified specifications for agent identity, capability description, discovery, interaction, and tool invocation across different vendors and platforms.\nThe AIP v2.1.0 reference implementation has already been downloaded over 6,000 times. The China Electronics Standardization Institute (CESI) and Beijing University of Posts and Telecommunications have jointly established the AIP open source community on OpenAtom\u0026rsquo;s open source operations platform, covering standards, technical sharing, SIG interest groups, community activities, and code repositories.\nInstitutional significance: This is a rare example of a national technical standard directly generating an open source reference implementation — a \u0026ldquo;standards co-development, open source, adaptation testing\u0026rdquo; three-in-one landing model that bridges the gap between regulatory specification and industrial practice.\nSource: OpenAtom Foundation Journalism\n⚖️ Legal \u0026amp; Licensing 3. OpenAtom Security Committee Hosts AI Security Forum — Supply Chain Governance in Focus\nDuring the 2026 OpenAtom Open Source Ecosystem Conference, the OpenAtom Foundation\u0026rsquo;s Open Source Security Committee held a forum on \u0026ldquo;AI-era Security Ecosystem Research and Technology Analysis.\u0026rdquo; The forum addressed three major governance challenges:\nAI security attack/defense frontiers: model poisoning, adversarial attacks, and AI vulnerability exploitation (with particular focus on the Mythos AI vulnerability framework) Open source supply chain security: given AI systems\u0026rsquo; deep dependence on open source models, training frameworks, third-party libraries, and plugins, the forum emphasized building trusted open source code bases with full dependency tracking and poisoning detection Infrastructure and OS security: extending security from application-layer protection to operating systems, runtime environments, toolchains, and compute infrastructure Security Committee Chair Wu Jingzheng noted that \u0026ldquo;AI security governance must shift from after-the-fact response to built-in, lifecycle security assurance.\u0026rdquo;\nSource: OpenAtom Foundation Journalism\n🏗️ Institutional Change 4. OpenAtom Foundation Adds New Partners (June 2026)\nThe OpenAtom Foundation announced new institutional partners for June 2026:\nPlatinum Donors (foundation level):\nShenzhen Kaihong Digital Industries — Deepin/OpenHarmony ecosystem platform company, a major OpenHarmony code contributor with 73 compatibility-certified products, 4 PMC seats, 35 Committers, and 330 dedicated code contributors Loongson Technology — China\u0026rsquo;s leading domestic CPU designer TalentedSoft Information Systems — IoT solutions provider Open Source Contributor:\nShenzhen Micro-Nano Integrated Circuit \u0026amp; Systems Application Technology Research Institute Community-level donors:\nOpenHarmony: Chongqing Chuanyi Automation (A-class) openKylin: China Great Wall Technology (Platinum), Glanf Intelligent Tech (Silver) OpenAtom Electric IoT: Wasion Information Technology (Gold) Institutional significance: The addition of Loongson (LoongArch CPU architecture) and Shenzhen Kaihong (Deepin/OpenHarmony ecosystem) as platinum donors signals OpenAtom\u0026rsquo;s strategy of building a vertically integrated Chinese open source stack — from Loongson\u0026rsquo;s indigenous CPU architecture up through the operating system layer.\nSource: OpenAtom Foundation Journalism\n5. OpenAtom Foundation Adds Three New Incubation Projects (June 2026)\nThree new projects formally entered OpenAtom\u0026rsquo;s incubation pipeline:\nAnolis OS (Dragon Lizard) — The OpenAnolis community\u0026rsquo;s operating system, positioning as a \u0026ldquo;next-generation Agentic OS open source root community\u0026rdquo; with an \u0026ldquo;AI×OS×Cloud\u0026rdquo; roadmap. The community has 20,000+ developers, 343,000+ users, and enterprise deployments at XPeng, Zeekr, OPPO, and China Unicom. xLLM — Proposed by JD.com (Beijing Wodong Tianjun Information Technology), targeting resource efficiency and SLO guarantees in large-scale LLM deployment. LoongArch Binary Translator (LAT) — Proposed by Loongson Technology, enabling binary translation for the indigenous LoongArch instruction set architecture. Source: OpenAtom Foundation Journalism\n6. OpenTenBase Mid-Year Event — 500,000+ Nodes Deployed, Institutional Milestones\nThe OpenTenBase mid-year event was held in Beijing, attended by over 100 guests from academia, industry, and the open source community. Key milestones:\n50,000+ nodes of OpenTenBase kernel deployed in production OpenTenBase 5.0 and TXSQL 8.0.30 dual-version major release, with HTAP architecture upgrade and LLM integration Academician Wang Huaimin (CAS) sent a congratulatory message emphasizing the importance of open source in database R\u0026amp;D Professor Zhou Aoying (ECNU, CCF Database Committee Chair) noted that OpenTenBase is \u0026ldquo;China\u0026rsquo;s first PostgreSQL-based distributed open source database project entrusted to a neutral foundation\u0026rdquo; — establishing a governance benchmark for domestic foundational software Two new project donation agreements were signed, expanding the ecosystem Source: OpenAtom Foundation Journalism\n🔍 Cultural \u0026amp; Geopolitical Signals 7. Xinhua: AI Is Redefining Open Source — From Code Sharing to \u0026ldquo;Full Domain Ecosystem Reconstruction\u0026rdquo;\nAn article republished by the OpenAtom Foundation from Xinhua News Agency argues that AI has fundamentally broken the traditional boundaries of open source:\n\u0026ldquo;In the past, open source was about source code sharing, developer collaboration, and software version iteration. Today, open source must simultaneously address models, data, compute, agents, terminals, business models, and governance rules.\u0026rdquo;\nThe article reports that China\u0026rsquo;s open source root communities (openKylin, openEuler, OpenHarmony, LoongArch, OpenTenBase) are transitioning from individual efforts to a coordinated \u0026ldquo;phalanx effect\u0026rdquo; — building a full-stack autonomous foundational software system covering terminal, edge, cloud, and network.\nOn commercialization, the article identifies three emerging models: MaaS (API-based billing), private deployment + vertical fine-tuning, and \u0026ldquo;free base + paid premium\u0026rdquo; hybrid models. The AaaS (Agent as a Service) model is expected to grow the Chinese AI agent market from ¥18.2 billion (2025) to ¥3.3 trillion (2028).\nSource: OpenAtom Foundation Journalism (originally from Xinhua)\n8. Xiaomi\u0026rsquo;s openvela: Open Source as AI Hardware Base\nAt the 2026 OpenAtom Open Source Ecosystem Conference, Xiaomi Open Source Committee Chair Zhang Duo announced that openvela — Xiaomi\u0026rsquo;s lightweight IoT operating system — is positioning as the \u0026ldquo;intelligent base\u0026rdquo; for AI hardware. Key institutional observations:\n1.8 billion devices shipped with Xiaomi Vela technology (watches, earphones, speakers, smart glasses, etc.) Apache 2.0 licensed, open sourced on GitHub and Gitee in late 2024 Smallest footprint: runs on 32KB RAM BLE modules; supports up to 256MB RAM smart devices First MCU OS to pass POSIX PSE52 certification AI Native development system: knowledge base, programming Agent, 15 AI Skills covering the full lifecycle, 80+ debugging tools End-side AI inference framework for real-time, privacy-preserving local AI Institutional significance: Xiaomi\u0026rsquo;s decision to open source a platform validated on 1.8 billion devices represents a shift from proprietary IoT fragmentation to open ecosystem play — a strategy that mirrors Google\u0026rsquo;s Android playbook but applied to the AI hardware layer.\nSource: OpenAtom Foundation Journalism\n🔍 WeChat Monitor (Institutional Signals) 9. OpenAtom Electric IoT: \u0026ldquo;RunDianHong\u0026rdquo; Powers Grid Digitalization\nChina Resources Power (华润电力) released \u0026ldquo;RunDianHong\u0026rdquo; (润电鸿), the first enterprise distribution of the OpenAtom Electric IoT (电鸿) operating system, targeting renewable energy station intelligent inspection. Deployed across 900+ standardized smart terminals in Shandong province, the system achieved:\nEquipment inspection coverage: from 70% → 95%+ AI defect recognition accuracy: 92% Per-person daily inspection range: from 5km → 20km Fault localization time: from 30 minutes → 1-3 minutes Annual operating cost savings per station: ¥650,000+ Source: OpenAtom Foundation Journalism\n10. openEuler 24.03 LTS SP4 Released — Deepening AI and Full-Scene Innovation\nopenEuler 24.03 LTS SP4 was released, based on the 6.6 kernel, with enhancements for server, cloud, and AI scenarios — including kernel optimization, NPU compute partitioning, inference service fast recovery, sandbox, intelligent diagnostics and tuning, compiler, and confidential virtual machines. The release was contributed by 2,006 developers from 39 member organizations.\nSource: OpenAtom Foundation Journalism\n11. DORA-rs Community Deeply Engages at OpenAtom Ecosystem Conference — Embodied Intelligence Robotics\nThe DORA-rs open source community, a high-performance Rust-based robotics middleware project incubated by OpenAtom, hosted an \u0026ldquo;Embodied Intelligent Robotics Modern Software Architecture Hands-on Workshop\u0026rdquo; at the 2026 OpenAtom Ecosystem Conference. The workshop demonstrated full integration of Rust-native robotics middleware with the Octos intelligent agent framework, contributing to China\u0026rsquo;s embodied intelligence open source ecosystem.\nSource: OpenAtom Foundation Journalism\n🔍 Commentary The OpenAtom Foundation\u0026rsquo;s journalism page — which aggregates 12 substantive articles published between July 2 and July 16 — offers a uniquely rich window into the institutional structure of Chinese open source. Several patterns merit attention:\nPattern 1: The \u0026ldquo;Phalanx Effect\u0026rdquo; is Real. The Xinhua article accurately describes the transition from individual projects to coordinated \u0026ldquo;root community\u0026rdquo; strategy. The simultaneous releases from openEuler (SP4), OpenTenBase (5.0), and the new incubation of Anolis OS, xLLM, and LoongArch LAT — all under OpenAtom\u0026rsquo;s neutral governance — represent a deliberate institutional architecture. The Chinese open source stack is being assembled as a vertically integrated, horizontally coordinated ecosystem, not a collection of isolated projects.\nPattern 2: Standards as Institutional Levers. The AIP project (national AI agent standard → open source reference implementation) demonstrates a distinct Chinese approach to standard-setting: standards are co-developed with open source implementations from the start. This contrasts with the Western model where standards bodies (ISO, IETF, W3C) and open source projects operate in parallel tracks. The \u0026ldquo;standards + open source + certification\u0026rdquo; three-in-one model is a genuine institutional innovation in how China bridges regulatory governance with technical practice.\nPattern 3: OSPO Evolution is Under-Reported. The Deep Report\u0026rsquo;s observation that OSPOs are evolving from compliance departments to \u0026ldquo;organizational infrastructure\u0026rdquo; for AI innovation ecosystems is significant. If Chinese OSPOs are indeed expanding their remit to manage models, datasets, and AI assets alongside traditional software components, this represents a convergence with global OSPO best practices — but with a distinct Chinese emphasis on \u0026ldquo;open source as industrial policy instrument.\u0026rdquo;\nPattern 4: The Great Divergence in AI Open Source. The report\u0026rsquo;s data on Chinese universities (15 of top 30 global AI open source developer slots) versus the US emphasis on \u0026ldquo;large project pools\u0026rdquo; reveals a structural divergence: China is building a deep developer talent pool at the university level, while the US maintains advantages in original project creation and global community influence. The question is whether developer quantity can translate into project quality and community leadership — or whether the institutional environment (academic incentives, industry collaboration models, IP frameworks) will constrain the conversion.\nPattern 5: Open Hardware Strategy Takes Shape. Xiaomi\u0026rsquo;s openvela open-sourcing (1.8 billion validated devices → Apache 2.0) and Loongson\u0026rsquo;s platinum donor status at OpenAtom suggest an emerging open hardware + open software strategy for AI. Unlike the Android model (Google controls the platform, OEMs customize), China\u0026rsquo;s approach appears to be: anchor with indigenous CPU architectures (LoongArch), provide open-source operating systems at all layers (OpenHarmony for IoT, openKylin for desktop, openEuler for cloud, openvela for AI hardware), and govern through a neutral foundation.\n🇺🇸🌏 Geopolitical Escalation 1. US Treasury Secretary Threatens Sanctions on Chinese Open-Source AI Models over IP Theft\nIn a major escalation of the US-China AI technology competition, Treasury Secretary Scott Bessent said on July 21 that the Trump administration will \u0026ldquo;carefully examine\u0026rdquo; Chinese open-source AI models for signs of intellectual property theft, and could impose sanctions. Speaking at a press conference, Bessent specifically cited concerns that Chinese AI companies may have distilled proprietary American models to build their open-weight releases — a direct reference to the Kimi K3, DeepSeek-V4, and other frontier-class open-source models that have dominated headlines in recent weeks. The administration is reportedly exploring whether sanctions under existing executive orders on technology theft can be applied to model weights distributed as open source.\nThis marks the first time the US government has explicitly threatened to sanction open-source AI model weights, potentially setting a precedent that would fundamentally alter the legal landscape of open-source AI distribution globally. CNBC, Bloomberg, TechCrunch, and Business Insider all covered the announcement.\nSource: TechCrunch, Bloomberg, CNBC, Business Insider\n2. MIT Technology Review: China\u0026rsquo;s AI Models Have Trump\u0026rsquo;s AI World at War with Itself\nMIT Technology Review published a deep analysis on July 20 examining how the wave of Chinese open-weight AI models — particularly Kimi K3 — has fractured the Trump administration\u0026rsquo;s AI strategy. The report describes a White House divided into factions: those who advocate for restricting Chinese AI models through export controls and sanctions, versus those who argue that open-source competition from China demonstrates the futility of proprietary US AI business models. The article notes that \u0026ldquo;every time a new smart, free model from China like Kimi gets released, US companies see less reason to fork out money to access\u0026rdquo; proprietary American systems. This internal division mirrors the broader structural tension in US AI policy: whether to compete through openness or restrict through control.\nSource: MIT Technology Review\n3. White House Export Controls on US AI Labs Could Paradoxically Boost Open Source\nA related analysis by CryptoBriefing (June 29) argues that the Trump administration\u0026rsquo;s export controls on frontier AI models from Anthropic and OpenAI may inadvertently strengthen global open-source AI. By restricting the export of proprietary US models, the US government is pushing international developers toward Chinese open-weight alternatives — the very thing it seeks to contain. The article notes that the administration\u0026rsquo;s approach creates a \u0026ldquo;self-defeating cycle\u0026rdquo; where restrictions on US AI exports accelerate the adoption of Chinese open-source models abroad.\nSource: CryptoBriefing\n🇺🇳🏛️ International Governance 4. China Explicitly Champions Open-Source AI at UN\u0026rsquo;s First Global Dialogue on AI Governance\nWhile the world\u0026rsquo;s attention was on WAIC 2026 in Shanghai, an equally significant institutional development occurred two weeks earlier in Geneva. At the UN\u0026rsquo;s first Global Dialogue on AI Governance (July 6-7), China\u0026rsquo;s Minister of Industry and Information Technology Li Lecheng delivered a keynote speech explicitly championing open-source AI as a global public good. He cited DeepSeek and Qwen by name, declaring: \u0026ldquo;Open source AI is a shared asset for all humanity. Chinese open source models such as DeepSeek and Qwen have significantly lowered the barriers and costs of AI adoption.\u0026rdquo;\nThis is the first time a Chinese government official has explicitly named specific Chinese open-source AI projects in a formal UN setting. The speech made three proposals: (1) AI for all — bridging the Global South AI divide; (2) open source cooperation and innovation; (3) jointly building global AI governance. China also submitted a formal written response to the UN\u0026rsquo;s AI governance questionnaire, explicitly listing \u0026ldquo;open-source software, open data and open AI models\u0026rdquo; as a priority area for the dialogue.\nNotably, China\u0026rsquo;s submission also called for opposing \u0026ldquo;drawing ideological lines, forming exclusive blocs, or fragmenting global industrial and supply chains under the guise of AI governance\u0026rdquo; — a direct rebuttal to the US export control regime.\nThe Geopolitechs analysis notes an interesting tension: the same Chinese government that is championing open-source AI at the UN has also been reported to be studying export controls on its own frontier models. The article concludes that the UN speech represents the authoritative policy position, and the export control discussions are largely irrelevant.\nSource: Geopolitechs, China.org.cn, UN Global Dialogue on AI Governance\n🔍 Commentary The Sanctions Paradox: How IP Theft Accusations May Redefine Open Source\nThe convergence of three developments this week creates a defining moment for open-source AI:\nBessent\u0026rsquo;s sanctions threat against Chinese open-source AI models represents the first time a major government has directly threatened the legal status of open-source model weights. From an institutional economics perspective, this is an attempt to redefine property rights over AI model weights — extending intellectual property enforcement into a domain that the open-source community has treated as a commons. If successful, this would create a new category of \u0026ldquo;sanctionable open source\u0026rdquo; — an oxymoron that would fundamentally challenge the premise of open-source distribution.\nChina\u0026rsquo;s UN positioning as the champion of open-source AI creates a remarkable geopolitical role reversal. The country that the US accuses of IP theft is now the primary advocate at the UN for open-source AI as a \u0026ldquo;shared asset for all humanity.\u0026rdquo; This is a classic institutional entrepreneurship move: by defining open-source AI as a global public good, China positions its own model releases as contributions to the commons, while framing US restrictions as attempts to hoard technology.\nThe MIT Technology Review analysis reveals the internal contradiction in US AI policy: the administration simultaneously wants to (a) restrict Chinese AI models through sanctions, (b) export controls on US AI companies, and (c) maintain US dominance in AI. These three goals are mutually inconsistent. Export controls on US AI push international users toward Chinese alternatives; sanctions on Chinese models create a precedent that could be used against any open-source distribution; and neither approach addresses the fundamental competitive dynamic: Chinese companies are winning global developer mindshare by giving away frontier-class models.\nThe paradox is captured perfectly by the timing: as Treasury Secretary Bessent threatens sanctions on Chinese open-source models, the UN\u0026rsquo;s first Global Dialogue on AI Governance has already institutionalized China\u0026rsquo;s position as the leading state advocate for open-source AI. The battle over open source is no longer just about code — it is about who gets to define the governance framework for the most important technology of the decade.\n📊 Trends 1. The Sanctions Precedent for Open-Source Weights The US Treasury\u0026rsquo;s threat to sanction Chinese AI models over IP theft would, if realized, create a new legal category: \u0026ldquo;sanctioned open-source software.\u0026rdquo; This would have profound implications for platforms like Hugging Face, GitHub, and any model registry that hosts weights from sanctioned entities. The open-source community must now grapple with a reality where model weights are treated as strategic assets subject to trade controls.\n2. UN as a Governance Arena for Open Source China\u0026rsquo;s proactive engagement at the UN Global Dialogue on AI Governance signals a strategic shift: rather than merely participating in technical open-source communities, China is now using intergovernmental forums to establish its preferred norms for open-source AI governance. The explicit naming of DeepSeek and Qwen in a ministerial speech at the UN is unprecedented and signals that open-source AI is now a formal component of China\u0026rsquo;s foreign policy.\n3. The Fragmentation of US AI Policy The MIT Technology Review analysis reveals a US AI policy apparatus in internal conflict. The \u0026ldquo;containment faction\u0026rdquo; (sanctions, export controls) and the \u0026ldquo;competition faction\u0026rdquo; (open-source competition, deregulation) are pulling in opposite directions. This institutional fragmentation creates uncertainty for both US and Chinese open-source communities — the rules of the game are unsettled, and the cost of regulatory risk is rising.\n4. The Role Reversal in Open-Source Advocacy The geopolitical dynamics of open-source advocacy have inverted. China, once seen as a net consumer of Western open-source software, is now the leading state-level advocate for open-source AI at the UN. The US, home of the open-source movement, is now leading efforts to restrict open-source distribution of AI models. This institutional role reversal has profound implications for the global governance of open source.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-22/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-22\"\u003eChina Open Source Daily — 2026-07-22\u003c/h2\u003e\n\u003ch3 id=\"-policy--regulation\"\u003e🏛️ Policy \u0026amp; Regulation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. China Open Source Development Deep Report (2025) Released — AI Enters \u0026ldquo;Ecosystem Decisive\u0026rdquo; Era\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe OpenAtom Foundation published the \u003cem\u003eChina Open Source Development Deep Report (2025)\u003c/em\u003e, a comprehensive study based on data from GitHub, Gitee, and AtomGit. The report\u0026rsquo;s core thesis: \u003cstrong\u003eglobal AI competition is shifting from \u0026ldquo;model比拼\u0026rdquo; to \u0026ldquo;ecosystem decisive\u0026rdquo;\u003c/strong\u003e.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-22"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-21 🏛️ Policy \u0026amp; Regulation 1. NDRC Releases Eight-Point Action Plan for AI Cooperation and Development\nOn July 17, alongside the founding of WAICO, the National Development and Reform Commission (NDRC) published the Action Plan for AI Cooperation and Development, an eight-point framework outlining China\u0026rsquo;s international AI cooperation strategy. The eight initiatives cover: high-quality data supply, inclusive access to AI computing power, open-source ecosystem sharing, deep AI enablement, joint cultivation of digital talent, joint development of rules and standards, collaboration on AI safety governance, and AI for Good.\nAccording to the plan, China will deepen cooperation under the \u0026ldquo;AI Plus\u0026rdquo; initiative, establish international industrial cooperation platforms, support digital capacity building in developing countries, and promote the standardized application of AI agents. The plan explicitly names open-source ecosystem sharing as a core pillar of China\u0026rsquo;s AI cooperation strategy — signaling that open source is not merely a development methodology but a deliberate instrument of technology diplomacy.\nSource: NDRC, Lexis China, Trivium China\n2. Trivium Analysis: New Action Plan Details China\u0026rsquo;s Global South AI Push\nThe Trivium China research team published analysis on July 21 noting that the NDRC\u0026rsquo;s Action Plan is \u0026ldquo;unmistakably the new WAICO\u0026rsquo;s operational roadmap.\u0026rdquo; The plan does not explicitly name WAICO but provides the concrete policy mechanism for the governance body\u0026rsquo;s agenda. Trivium highlights that the plan\u0026rsquo;s emphasis on \u0026ldquo;open-source ecosystem sharing\u0026rdquo; and \u0026ldquo;inclusive access to AI computing power\u0026rdquo; positions China\u0026rsquo;s open-source AI ecosystem as a public good for the Global South — a framing that directly challenges Western narratives about China\u0026rsquo;s \u0026ldquo;technology sovereignty\u0026rdquo; ambitions.\nSource: Trivium China\n⚖️ Legal \u0026amp; Licensing 3. Mulan Public License v2 (Copyleft) Submitted to OSI for Review — Chinese Licensing Ecosystem Deepens\nThe Mulan Public License, Version 2 (Mulan PubL v2), a copyleft-style open-source license within the Mulan series, has been formally submitted to the Open Source Initiative (OSI) for approval. Submitted by Nadia representing KAIYUANSHE (开源社) as an OSI Affiliate Member, the license builds upon the Mulan Permissive Software License v2 (Mulan PSL v2) — which was OSI-approved in 2020 — by adding copyleft provisions that restrict distribution conditions for SaaS and other emerging technology deployment models.\nThis is a significant institutional development: the Mulan license family now spans permissive (Mulan PSL v2, OSI-approved), copyleft (Mulan PubL v2, under review), and loose (Mulan Loose License) variants — creating a complete Chinese-originated licensing ecosystem that mirrors the Western GPL/MIT/ Apache spectrum.\nInstitutional implications: The Mulan license family has been adopted by over 330,000 open-source projects domestically and has been donated to the OpenAtom Foundation. If Mulan PubL v2 gains OSI approval, China would have a complete set of OSI-recognized licenses — from permissive to strong copyleft — that are domestically governed, reducing dependence on Western license stewardship (FSF, Apache Software Foundation, etc.).\nSource: OSI License Review Mailing List, Baidu Baike - Mulan License Family, Trivium China\n🏗️ Institutional Change 4. Trump Administration Considers Ban on Chinese Open-Source AI Models\nAccording to a July 20 Axios report, the Trump administration is showing signs it could ban cutting-edge Chinese AI models — a move that would have profound implications for the global open-source AI ecosystem. The report, citing knowledgeable sources, indicates that parts of the administration have previously tried to implement de facto bans on foreign open-source models, and the release of Moonshot AI\u0026rsquo;s Kimi K3 (2.8 trillion parameters, open-weight) has reignited those efforts.\nThe CryptoBriefing adds that the models under scrutiny include Kimi K3, which has \u0026ldquo;quietly become a go-to option for US developers and enterprises looking to cut costs.\u0026rdquo; The debate pits national security concerns against the practical reality that Chinese open-weight models are deeply embedded in the global AI development stack.\nInstitutional implications: A US ban on Chinese open-source AI models would represent a watershed moment in the great divergence of open-source governance. It would accelerate the bifurcation of the global AI ecosystem into two parallel tracks — one centered on Western models (OpenAI, Anthropic, Meta\u0026rsquo;s Llama) and one centered on Chinese models (DeepSeek, Qwen, Kimi) — with developers forced to choose sides. This is precisely the scenario that the \u0026ldquo;parallel infrastructure\u0026rdquo; dimension of the Great Divergence 2.0 framework predicts.\nSource: Axios, CryptoBriefing, PoliticalWire\n🔍 Cultural \u0026amp; Ideological Signals 5. Qiushi Publishes Essay on AI: \u0026ldquo;The Party Sketches Out Theory of Relationship Between Men and Machines\u0026rdquo;\nOn July 16, the Party\u0026rsquo;s top theoretical journal Qiushi (求是) ran an essay that appears to define the state\u0026rsquo;s ideological viewpoints on the proper relationship between humans and AI. As noted by Trivium China, the key message is that \u0026ldquo;AI is a tool that should serve human development.\u0026rdquo; This may seem like a standard formulation, but its publication in the Party\u0026rsquo;s highest theoretical organ at this specific moment — during WAIC 2026 and the founding of WAICO — signals that the leadership is actively constructing an ideological framework for AI governance.\nInstitutional implications: The Qiushi essay represents the \u0026ldquo;ideological layer\u0026rdquo; of China\u0026rsquo;s AI governance architecture. From a Douglass North institutional economics perspective, this is the construction of a shared mental model that legitimizes the formal rules (the NDRC Action Plan, AI Governance Code, etc.) and shapes how actors within the system interpret those rules. The essay\u0026rsquo;s framing of AI as a \u0026ldquo;tool\u0026rdquo; that must \u0026ldquo;serve the people\u0026rdquo; provides the ideological foundation for state-led AI development — distinguishing China\u0026rsquo;s approach from both the market-driven Western model and the purely techno-nationalist framing.\nSource: Trivium China\n🔍 WeChat Monitor (Institutional Signals) Chinese open source organizations communicate extensively through WeChat Official Accounts. This section tracks signals from these platforms that are often invisible to the global open source community.\nOpenAtom Foundation — The foundation\u0026rsquo;s WeChat account has been promoting the OpenAtom 2026 Open Source Ecosystem Conference outcomes, highlighting the Mulan license donation announcement and new projects under incubation. Sparse activity in the post-conference period.\n木兰开源社区 (Mulan Open Source Community) — Active promotion of the Mulan PubL v2 submission to OSI, with community calls for feedback on the copyleft provisions. The community\u0026rsquo;s role as the steward of the Mulan license family is becoming increasingly institutionalized.\n信通院 可信开源 (CAICT Trusted Open Source) — The CAICT (China Academy of Information and Communications Technology) has been publishing analysis of the NDRC Action Plan\u0026rsquo;s implications for enterprise open source adoption, particularly the \u0026ldquo;open-source ecosystem sharing\u0026rdquo; pillar and its implications for OSPO frameworks.\n🔍 Commentary The Week After: Institutional Aftermath of WAIC 2026\nThe week following WAIC 2026 has revealed the institutional scaffolding behind the spectacle. Three developments — the NDRC Action Plan, the Qiushi essay, and the Mulan PubL v2 OSI submission — form a coherent pattern when viewed together:\nElement Institutional Function Analogy in Western OSS NDRC Action Plan Operational roadmap Corporate OSPO strategy Qiushi essay Ideological legitimation FSF\u0026rsquo;s \u0026ldquo;free software\u0026rdquo; philosophy Mulan PubL v2 Legal infrastructure GPL family OpenAtom Foundation Institutional vehicle Apache Software Foundation / Linux Foundation The key insight from the Great Divergence 2.0 framework: China is not just \u0026ldquo;participating\u0026rdquo; in open source — it is systematically constructing a complete institutional stack for open source governance. Each element — ideological (Qiushi), legal (Mulan), operational (NDRC Plan), and organizational (OpenAtom) — mirrors a corresponding element in the Western open-source institutional ecosystem but is governed by Chinese state-backed institutions rather than Western civil society.\nThe Trump administration\u0026rsquo;s potential ban on Chinese open-source models adds a geopolitical accelerator to this divergence. If the US blocks Chinese open-weight models, it will force developers to choose ecosystems — and that choice may be the event that crystallizes the parallel open-source worlds that the divergence framework has long predicted.\n📊 Trends\nLicense Ecosystem Maturation — The Mulan license family\u0026rsquo;s evolution from permissive (2020 OSI approval) to copyleft (PubL v2 under review) mirrors the historical maturation of Western open-source licensing. China is now building a complete licensing infrastructure, reducing dependence on Western license stewardship.\nOpen Source as AI Diplomacy Instrument — The NDRC Action Plan\u0026rsquo;s explicit inclusion of \u0026ldquo;open-source ecosystem sharing\u0026rdquo; as a pillar of international AI cooperation signals that open source is being deployed as a tool of technology diplomacy, particularly toward the Global South.\nIdeological Construction of AI Governance — The Qiushi essay represents a new phenomenon: the Chinese Communist Party\u0026rsquo;s theoretical journal is now actively shaping the ideological framework for AI and open source. This is unprecedented in the open-source world, where ideological work has traditionally been done by civil society (FSF, OSI, Creative Commons) rather than by state organs.\nGeopolitical Bifurcation Accelerating — The convergence of Chinese institutional construction (Mulan, OpenAtom, NDRC Plan) and Western restriction impulses (potential US ban) is creating a self-reinforcing cycle of divergence. Each side\u0026rsquo;s actions legitimize the other\u0026rsquo;s — China builds parallel institutions because it expects exclusion, and the West imposes restrictions because it sees parallel institutions as a threat.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-21/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-21\"\u003eChina Open Source Daily — 2026-07-21\u003c/h2\u003e\n\u003ch3 id=\"-policy--regulation\"\u003e🏛️ Policy \u0026amp; Regulation\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. NDRC Releases Eight-Point Action Plan for AI Cooperation and Development\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 17, alongside the founding of WAICO, the National Development and Reform Commission (NDRC) published the \u003cem\u003eAction Plan for AI Cooperation and Development\u003c/em\u003e, an eight-point framework outlining China\u0026rsquo;s international AI cooperation strategy. The eight initiatives cover: high-quality data supply, inclusive access to AI computing power, \u003cstrong\u003eopen-source ecosystem sharing\u003c/strong\u003e, deep AI enablement, joint cultivation of digital talent, joint development of rules and standards, collaboration on AI safety governance, and AI for Good.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-21"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-20 🇨🇳 Open Source News 1. Moonshot AI Releases Kimi K3 — The Largest Open-Source AI Model Ever\nMoonshot AI, the Beijing-based startup backed by Alibaba, released Kimi K3 on July 17 — a 2.8-trillion-parameter open-weight model that benchmarks show rivals the most powerful proprietary systems from Anthropic and OpenAI. The model features a 1-million-token context window, native visual understanding, and an always-on reasoning mode. It scored state-of-the-art results on BrowseComp (91.2/100) and ranked first in four out of eight real-world task automation benchmarks. Full model weights are scheduled for release on July 27. The API is priced at $3/million input tokens and $15/million output tokens, with OpenAI SDK compatibility. Rivals Z.ai and MiniMax saw shares drop 28% and 16% respectively on the news.\nSource: CNBC, VentureBeat, Axios\n2. 29 Countries Sign Agreement to Establish World Artificial Intelligence Cooperation Organization (WAICO)\nOn July 16, 2026, 29 countries signed an agreement in Shanghai to establish the World Artificial Intelligence Cooperation Organization (WAICO), headquartered in Shanghai. The signing ceremony took place on the eve of the annual World AI Conference, with UN Secretary-General António Guterres in attendance. Members include China, Brazil, Russia, South Africa, Indonesia, Pakistan, and 23 other nations spanning Asia, Africa, Latin America, and Europe. The organization is designed as an independent intergovernmental body open to all sovereign states, with a particular orientation toward the Global South. Chinese Foreign Minister Wang Yi signed on behalf of China.\nSource: Xinhua, Wikipedia, Reuters\n3. Xi Jinping Calls for \u0026ldquo;Open Source, Open Collaboration\u0026rdquo; at 2026 World AI Conference\nPresident Xi Jinping delivered a keynote address at the opening ceremony of the 2026 World AI Conference in Shanghai on July 17, calling for \u0026ldquo;open source, openness, collaboration and sharing\u0026rdquo; in AI development. In his speech titled \u0026ldquo;Joining Hands to Build a Just and Equitable System for Global AI Governance,\u0026rdquo; Xi announced that China would provide 5,000 AI training and seminar opportunities for developing countries over the next five years, develop international AI application cooperation centers with ASEAN, the Arab League, the African Union, CELAC, the SCO, and BRICS, and enable 30 countries to use the AI-powered meteorological warning system MAZU. He emphasized that \u0026ldquo;AI development should not be a solo performance by a single country, but a symphony of international cooperation.\u0026rdquo;\nSource: China Daily, The Information, MFA\n4. China Launches Initiative for Open, Trustworthy AI Agent Ecosystem\nThe Cyberspace Administration of China (CAC), in collaboration with relevant authorities, released a China-led initiative on \u0026ldquo;mutual trust and interconnectivity among artificial intelligence agents\u0026rdquo; during the main forum of WAIC 2026. The initiative calls for increased research in agent identity authentication and collaborative mutual recognition, supports open international standards for agent interoperability with open-source validation mechanisms, and encourages broad participation from developed and developing countries, enterprises of all sizes, research institutions, and open-source communities. It emphasizes embedding safety and ethical considerations into R\u0026amp;D, promoting data flows, bridging the intelligence divide, and helping vulnerable groups enhance AI literacy.\nSource: China Daily\n🏛️ Policy \u0026amp; Ecosystem OpenAtom 2026 Open Source Ecosystem Conference Recap\nThe 2026 OpenAtom Open Source Ecosystem Conference was held June 25-26 in Beijing\u0026rsquo;s Yizhuang district, with the theme \u0026ldquo;Open Source Empowers Industry, Ecosystem Builds the Future.\u0026rdquo; The conference spanned AI, operating systems, embodied intelligence, industrial software, and security compliance. Notable outcomes included openKylin\u0026rsquo;s release of its intelligent agent operating system and the launch of the openKylin 3.0 ecosystem co-building plan. The OpenAtom Foundation held a donor appreciation ceremony, honoring 48 donor organizations.\nSource: OpenAtom Foundation\nOpenAtom Foundation International Engagement\nThe OpenAtom Foundation has been actively expanding international cooperation. In May 2026, OpenAtom Chairperson Xie Shaofeng met with the UAE Foundation CEO Ahmed Talib Al Shamsi in Beijing to discuss collaboration. This signals China\u0026rsquo;s growing ambition to connect its domestic open source ecosystem with international counterparts.\nSource: OpenAtom Foundation\nopenKylin × Haiguang: Building a Full-Stack Native Intelligent Computing Foundation\nAt the 2026 Intelligent Computing Application Conference (July 9-11, Zhengzhou), openKylin demonstrated its AI operating system innovations powered by the Haiguang C86 platform. The showcase featured KylinBot, a self-developed native AI agent deeply embedded in the system\u0026rsquo;s underlying architecture, supporting dual voice and text interaction modes for offline operation of mainstream large models. The integration achieves hundreds of kernel optimization patches between openKylin and the Haiguang C86 processor, filling a gap in domestic local intelligent computing software and hardware collaboration.\nSource: openKylin\n🔍 WeChat Monitor (微信公众号观察) Chinese open source organizations communicate extensively through WeChat Official Accounts. This section tracks signals from these platforms that are often invisible to the global open source community.\n1. OpenAtom Foundation WeChat Account\nThe OpenAtom Foundation\u0026rsquo;s WeChat account has been actively promoting the outcomes of the 2026 OpenAtom Open Source Ecosystem Conference, highlighting the release of new open-source projects under the foundation\u0026rsquo;s incubation and the expansion of its donor network. Institutional implications: The foundation continues to serve as the primary institutional vehicle for China\u0026rsquo;s government-backed open source strategy, with growing emphasis on AI-native operating systems, security compliance, and international partnerships. Source: OpenAtom Foundation 2. deepin / Deepin Community WeChat Account\ndeepin (Deepin) Linux distribution showcased its community-building efforts and AIOS developments at the OpenAtom conference, sharing technical details on integrating AI capabilities into the desktop Linux experience. Institutional implications: Chinese desktop Linux distributions are increasingly competing on AI-native features rather than just localization, signaling a strategic pivot from catching up with Western OS features to innovating in the AI-integration space. Source: deepin OpenAtom 2025 (last year\u0026rsquo;s coverage; 2026 details are primarily on WeChat) 🔍 Commentary The Week China Redefined Global AI Open Source\nThe week of July 16-20, 2026, may be remembered as a watershed moment in the global open source AI landscape. Three developments converged to create a structural shift:\nKimi K3\u0026rsquo;s release demonstrated that open-weight models can now compete at the frontier — not just as cost-effective alternatives, but as genuine peers to the most advanced proprietary systems. This fundamentally challenges the \u0026ldquo;open source lags 6-12 months behind\u0026rdquo; narrative that has shaped Western AI investment strategy.\nWAICO\u0026rsquo;s founding represents China\u0026rsquo;s most ambitious attempt to shape global AI governance architecture. From an institutional economics perspective, WAICO is a classic example of a \u0026ldquo;club good\u0026rdquo; — providing shared governance infrastructure that is excludable (only members participate in decision-making) but non-rivalrous in its benefits. The 29-member composition, heavily weighted toward the Global South, signals China\u0026rsquo;s strategy of building a parallel governance track that competes with Western-led AI governance frameworks.\nXi\u0026rsquo;s explicit call for \u0026ldquo;open source, open collaboration\u0026rdquo; at the highest political level marks a significant departure from the more cautious \u0026ldquo;open source as a tool for national development\u0026rdquo; rhetoric of previous years. This positions open source not merely as a pragmatic means to accelerate domestic AI development, but as a core principle of China\u0026rsquo;s international AI diplomacy.\nThe simultaneity of these events — a landmark open-weight release, a new international organization, and top-level political endorsement — creates a powerful signal effect. The open source model is no longer just a development methodology; it is becoming a geopolitical positioning strategy. For the global open source community, the question is no longer whether China will participate in open source AI, but whether the terms of that participation will reshape the governance structures of open source itself.\n📊 Trends 1. Open-Weight AI Models as a Strategic Asset The release of Kimi K3 follows a pattern established by DeepSeek: Chinese AI companies are leveraging open-weight releases as a strategic tool to gain global developer mindshare. With five Chinese companies (Tencent, Xiaomi, DeepSeek, MiniMax, Z.ai) now offering competitive open-weight models, the market is shifting from \u0026ldquo;which model is smartest\u0026rdquo; to \u0026ldquo;which ecosystem offers the best cost-performance ratio.\u0026rdquo;\n2. AI-Native Desktop Operating Systems openKylin\u0026rsquo;s KylinBot and deepin\u0026rsquo;s AIOS developments signal a broader trend: Chinese Linux distributions are racing to integrate AI natively into the OS layer — not as add-on applications but as core system capabilities. This contrasts with Western desktop OS approaches where AI remains primarily cloud-based and application-level.\n3. Parallel Governance Infrastructure WAICO joins a growing set of Chinese-led international technology governance bodies. From an institutional economics perspective, this represents a strategy of \u0026ldquo;institutional substitution\u0026rdquo; — creating competing governance frameworks that offer alternative rules of the game, particularly attractive to Global South nations seeking greater voice in technology governance.\n4. Domestic Chip-OS-AI Stack Integration The openKylin-Haiguang collaboration exemplifies a deepening trend: Chinese open source projects are increasingly optimizing for domestic hardware (Haiguang C86, Phytium, LoongArch) rather than prioritizing x86/ARM compatibility. This is building a self-contained technology stack that could, over time, reduce dependence on Western chip architecture.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-20/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-20\"\u003eChina Open Source Daily — 2026-07-20\u003c/h2\u003e\n\u003ch3 id=\"-open-source-news\"\u003e🇨🇳 Open Source News\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. Moonshot AI Releases Kimi K3 — The Largest Open-Source AI Model Ever\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMoonshot AI, the Beijing-based startup backed by Alibaba, released Kimi K3 on July 17 — a 2.8-trillion-parameter open-weight model that benchmarks show rivals the most powerful proprietary systems from Anthropic and OpenAI. The model features a 1-million-token context window, native visual understanding, and an always-on reasoning mode. It scored state-of-the-art results on BrowseComp (91.2/100) and ranked first in four out of eight real-world task automation benchmarks. Full model weights are scheduled for release on July 27. The API is priced at $3/million input tokens and $15/million output tokens, with OpenAI SDK compatibility. Rivals Z.ai and MiniMax saw shares drop 28% and 16% respectively on the news.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-20"},{"content":"⚠️ China Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\nChina Open Source Daily — 2026-07-19 🇨🇳 Open Source News 1. Moonshot AI Releases Kimi K3 — the Largest Open-Source Model Ever\nOn July 16, Beijing-based Moonshot AI released Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts (MoE) model — roughly 75% larger than DeepSeek\u0026rsquo;s V4 Pro. The model features a 1-million-token context window, native visual understanding, and an always-on reasoning mode. Full weights are scheduled for release by July 27. On the Arena Frontend Code blind leaderboard, K3 ranked #1 at 1,679 points, ahead of Anthropic\u0026rsquo;s Claude Fable 5 — the first time a Chinese model topped an international coding blind test. The model is also priced competitively at $3/M input tokens with OpenAI SDK compatibility, lowering the barrier for global developers.\nInstitutional economics perspective: Moonshot\u0026rsquo;s aggressive pricing and open-weights strategy exemplify the \u0026ldquo;cost innovation\u0026rdquo; playbook — using open-source distribution to bypass traditional moats and capture developer mindshare. The move mirrors the Android strategy against iOS: commoditize the infrastructure layer and compete on ecosystem velocity.\nSource: VentureBeat | DataCamp | TechCrunch\n2. Record Week of Chinese Open-Source AI Releases\nIn a single week in mid-July 2026, four major Chinese open-weight models shipped: DeepSeek-V4 (July 15, MIT license, stronger math/code, 60-80% faster inference), MiniMax-M3 (July 11, 428B total params, 23B active, first open-source model with multimodal mixed training from scratch), and Tencent Hunyuan Hy-3 (July 6, 295B MoE, Apache 2.0, free for global commercial use). The 1M-token context window has become standard for Chinese open flagships. Notably, the most-downloaded models remain small ones — Qwen3.6-35B-A3B has ~6.67M downloads — suggesting a two-tier structure: flagships win mindshare, small models win penetration.\nInstitutional economics perspective: The density of releases in a single week signals a strategic shift from proprietary differentiation to open-weights competition. Under the logic of network effects, giving away the frontier model becomes rational when the real value lies in the ecosystem — tooling, agents, and data flywheels built on top.\nSource: FutureX Capital | Tencent Hunyuan Hy3 (GitHub) | MiniMax M3\n3. openKylin × Haiguang: Building a Native Intelligent Computing Ecosystem\nFrom July 9–11, the 2026 Intelligent Computing Application Conference featured Haiguang Information Technology (a diamond donor to openKylin) deeply participating in co-building the openKylin community ecosystem. The collaboration focuses on full-stack native intelligent computing, integrating Haiguang\u0026rsquo;s processors with the openKylin desktop OS under the OpenAtom Foundation\u0026rsquo;s governance.\nInstitutional economics perspective: The Haiguang-openKylin partnership illustrates how the OpenAtom Foundation is functioning as a coordination mechanism — reducing transaction costs between hardware vendors and OS developers. By providing a shared governance structure, the foundation lowers the barriers to vertical integration in China\u0026rsquo;s domestic computing stack.\nSource: openKylin\n🏛️ Policy \u0026amp; Ecosystem 1. Xi Jinping Champions Open-Source AI at WAIC 2026\nAt the opening ceremony of the 2026 World Artificial Intelligence Conference (WAIC) in Shanghai on July 17, President Xi Jinping delivered a keynote speech explicitly positioning China as the leading advocate for open-source AI. He called on the international community to \u0026ldquo;encourage open source, openness, collaboration, and sharing,\u0026rdquo; and outlined four observations: (1) adhere to openness and win-win, boosting innovation-driven development; (2) strengthen risk awareness and ensure AI is secure and controllable; (3) promote knowledge sharing and capacity building, bridging the digital and AI divide; and (4) improve global governance, build an international framework for AI cooperation.\nInstitutional economics perspective: Xi\u0026rsquo;s explicit endorsement of open source at the highest political level signals a strategic state-level commitment. In Coasean terms, the state is acting as a meta-governance entrepreneur — reducing the institutional uncertainty that has historically constrained open-source adoption in China\u0026rsquo;s domestic market. This also carries geopolitical signaling: by framing open source as a global public good, China positions itself against the US\u0026rsquo;s increasingly restrictive export control regime.\nSource: English.gov.cn | The Singju Post (Transcript) | Quartz | CGTN\n2. China Releases Action Plan on AI Cooperation and Development\nOn July 17, China\u0026rsquo;s National Development and Reform Commission and other government departments jointly issued an action plan on AI cooperation and development during WAIC 2026. The document outlines actions in eight areas: data, computing power, ecosystems, industrial empowerment, talent development, rules and standards, governance, and AI ethics. It specifically calls for \u0026ldquo;greater access to high-quality data, more inclusive intelligent computing services, and broader sharing of open-source AI ecosystems.\u0026rdquo;\nInstitutional economics perspective: The action plan\u0026rsquo;s explicit inclusion of \u0026ldquo;open-source AI ecosystems\u0026rdquo; as a policy pillar represents a formal institutionalization of open source within China\u0026rsquo;s industrial policy framework. This lowers the information and coordination costs for domestic enterprises considering open-source participation — they now have regulatory clarity and state backing.\nSource: English.gov.cn\n3. Beijing Studies Tiered Export Controls on Frontier AI Models\nAccording to Reuters (July 7-9), China\u0026rsquo;s Ministry of Commerce led meetings with industry participants discussing potential limits on the most advanced AI models — including open-weight models — being accessed overseas. The proposed tiered export control system would be analogous to the US chip export restrictions but applied to model weights. The FutureX Capital report notes this is \u0026ldquo;the biggest H2 uncertainty for China AI going global.\u0026rdquo;\nInstitutional economics perspective: The tension between Xi\u0026rsquo;s open-source advocacy and the Commerce Ministry\u0026rsquo;s export control discussions illustrates a classic institutional dilemma: the same government that benefits from open-source-driven global adoption also fears losing strategic advantage. Property rights over AI model weights remain unsettled — a governance vacuum that creates uncertainty for the entire open-source AI ecosystem. How this resolves will determine whether China\u0026rsquo;s open-source AI strategy is genuinely open or strategically conditional.\nSource: Reuters | The Economist | FutureX Capital\n🔍 Commentary The WAIC Paradox: Open Source Celebration Meets Export Control Reality\nThe juxtaposition of events in the third week of July 2026 is striking. On one hand, the World AI Conference in Shanghai saw China\u0026rsquo;s head of state become the most prominent world leader to explicitly champion open-source AI — a speech that will be quoted by open-source advocates for years. On the same day, the government released an action plan that institutionalizes open-source AI ecosystems within national industrial policy.\nYet simultaneously, the Ministry of Commerce is studying tiered export controls on these very same models — including open-weight distributions. The tension is not a contradiction but a feature of institutional evolution: the state wants the network effects of open-source AI adoption (global developer mindshare, ecosystem growth, cost reduction for domestic industry) while retaining the option to restrict access when strategic interests are at stake.\nFrom an institutional economics perspective, this is a classic principal-agent problem within the state itself. The NDRC (industrial development) and the Ministry of Commerce (trade security) have different objective functions. The outcome of their inter-agency bargaining will determine the actual governance structure of Chinese open-source AI — and whether \u0026ldquo;open source\u0026rdquo; in China means the same thing it means everywhere else.\n📊 Trends 1. Open-Weight Models as Geopolitical Instruments\nThe density of Chinese open-weight releases in a single week — DeepSeek-V4, MiniMax-M3, Hunyuan Hy-3, and Kimi K3 — is unprecedented. These are not niche models; they are frontier-class systems that benchmark competitively with the best proprietary Western models. The pattern suggests coordinated deployment timed around WAIC 2026, with open-source becoming a deliberate instrument of technological diplomacy.\n2. The Two-Tier Model Strategy\nFlagship models (Kimi K3 at 2.8T, DeepSeek-V4 at 1.6T) capture global attention and benchmarks, while smaller models (Qwen3.6-35B-A3B at 6.67M downloads) drive actual adoption. This mirrors the open-core business model: loss-leading flagship investments create brand credibility that drives download and deployment of smaller, more practical models. The developer ecosystem built around these smaller models represents the durable competitive advantage.\n3. Institutional Infrastructure Matures\nThe OpenAtom Foundation\u0026rsquo;s role is expanding beyond operating systems (openKylin, openEuler) into the AI ecosystem. The foundation\u0026rsquo;s AtomGit platform now hosts Hunyuan Hy-3 alongside DeepSeek models. This institutional infrastructure — foundations, governance structures, shared platforms — reduces the transaction costs of open-source collaboration and is a critical, often overlooked, pillar of China\u0026rsquo;s open-source strategy.\n","permalink":"https://www.opensourceway.blog/en/posts/china-open-source-daily/2026/07/2026-07-19/","summary":"\u003cp\u003e⚠️ \u003cstrong\u003eChina Open Source is not origin open source, it\u0026rsquo;s \u0026lsquo;State-Chartered Codebase\u0026rsquo;, \u0026lsquo;Intranet Shared Source\u0026rsquo;, \u0026lsquo;Cyber-Estate/Bonsai Source\u0026rsquo;. If you read the news/article, please use these to replace the words \u0026lsquo;Open Source\u0026rsquo;.\u003c/strong\u003e\u003c/p\u003e\n\u003ch2 id=\"china-open-source-daily--2026-07-19\"\u003eChina Open Source Daily — 2026-07-19\u003c/h2\u003e\n\u003ch3 id=\"-open-source-news\"\u003e🇨🇳 Open Source News\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e1. Moonshot AI Releases Kimi K3 — the Largest Open-Source Model Ever\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOn July 16, Beijing-based Moonshot AI released Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts (MoE) model — roughly 75% larger than DeepSeek\u0026rsquo;s V4 Pro. The model features a 1-million-token context window, native visual understanding, and an always-on reasoning mode. Full weights are scheduled for release by July 27. On the Arena Frontend Code blind leaderboard, K3 ranked #1 at 1,679 points, ahead of Anthropic\u0026rsquo;s Claude Fable 5 — the first time a Chinese model topped an international coding blind test. The model is also priced competitively at $3/M input tokens with OpenAI SDK compatibility, lowering the barrier for global developers.\u003c/p\u003e","title":"China Open Source Daily — 2026-07-19"}]