⚠️ Editorial note: The open source ecosystem in China operates under a distinct institutional framework — characterized by state-led initiatives, intranet-like boundaries, and top-down governance. Readers should be aware that this context differs from the community-driven open source model common in other regions. The term “open source” as used in Chinese media may refer to practices that diverge from the conventional definition.

China Open Source Daily — 2026-09-06

🏗️ Institutional Change — MIIT’s AI SME Startup Support Plan: Five Administrative-Document Moves on the Same Unsolved Object as the Prior Four Briefings

1. MIIT Office Document 工信厅企业〔2026〕31号 — “AI SME Startup Support Plan (2026-2028)”, issued August 29, 2026, publicly released September 4, 2026

2. Sina Finance / 郭冀川 (September 5, 2026) — “工信部发布《人工智能中小企业创业支持计划(2026—2028年)》 三年培育超万家AI科创中小企业”

3. Tencent News (September 4, 2026) — “五问+一图，读懂《人工智能中小企业创业支持计划（2026—2028年）》” (MIIT SME Bureau Q&A on the plan)

4. DoNews / 智通财经 (September 4, 2026) — “工信部启动人工智能中小企业创业支持计划”

MIIT’s office issued on August 29, 2026 — and publicly released on September 4, 2026 — a 15-item three-year programme: the AI SME Startup Support Plan (人工智能中小企业创业支持计划) (document reference 工信厅企业〔2026〕31号). The plan was signed by the Ministry’s SME Bureau, per the Sina Finance report and the Tencent News Q&A with the MIIT SME Bureau’s responsible official. The 15 items are grouped into four blocks: (i) 强化创业要素供给 (supply of entrepreneurial factors) — items 1–3; (ii) 加强创业主体培育 (cultivation of entrepreneurial subjects) — items 4–8; (iii) 深化开源生态赋能 (deepening open-source ecosystem enablement) — items 9–11; (iv) 完善创业服务保障 (improving entrepreneurial-service guarantees) — items 12–15.

The three-year quantitative targets, documented in the plan’s own text: new-tech-style SMEs ≥ 10,000; 专精特新 “little giant” SMEs > 2,000; 10 AI-focused technology-incubators; 10 national SME-service demonstration platforms; 10 national-level SME feature-industry clusters.

Five institutionally-new administrative moves

Reading the 15 items in the order they appear in the plan, four of them make administrative-document moves that prior briefings did not document — and each of the four moves is on the same unsolved institutional object that the prior four briefings (September 2, 3, 5) have now documented across five institutional surfaces: the governance layer of China’s open-weight ecosystem.

Move 1 — Item 9 formally names AtomGit as a “national-level AI open-source community” (国家级人工智能开源社区).

The plan’s item 9 reads: “推进开源创业生态建设。支持国家级人工智能开源社区（AtomGit）发展壮大，加强面向人工智能创业企业的优质开源资源供给和服务支撑。鼓励人工智能创业企业积极参与开源生态建设，贡献开源模型、工具、数据集等资源。” (Advance the construction of an open-source entrepreneurial ecosystem. Support the growth of the national-level AI open-source community (AtomGit), strengthen the supply of quality open-source resources and service support to AI entrepreneurial enterprises. Encourage AI entrepreneurial enterprises to actively participate in the open-source ecosystem construction, contributing open-source models, tools, datasets, and other resources.)

This is the first administrative-document designation of AtomGit at the national-community level. Prior briefings documented AtomGit at the code-hosting layer (August 19–22, as a Huawei-adjacent alternative to GitHub) and at the Great Divergence 2.0 State-Chartered Codebase surface (as the localised code-hosting pole), but none documented AtomGit’s formal national-community designation. From an institutional economics standpoint, this code-hosting-platform / national-community-designation pairing is a structural finding: AtomGit’s institutional identity is now formally an administrative-object (national-community-designated) rather than a commercial-platform identity — a placement that moves AtomGit from a vendor product into the same administrative tier as Xie Lan’s 中国社会科学报 statecraft framing (August 31) and the tiered-governance apparatus (Reuters / Jamestown, July 7 / August 13).

Move 2 — Item 4 explicitly names “open-source community core contributors” as an entrepreneurial subject with an encouraged commercialization path.

The plan’s item 4 reads: “鼓励各地对依托智能工具开展敏捷创业的’一人公司’、超级个体等微型主体给予包容支持。加快培育’智能原生’企业，支持开源社区核心贡献者、智能体开发者等新兴群体探索商业化路径，支持一批具备高成长潜力的创业新锐。” (Encourage local governments to give inclusive support to micro-subjects such as “one-person companies” and super-individuals that rely on intelligent tools for agile entrepreneurship. Accelerate the cultivation of “intelligent-native” enterprises, support emerging groups such as open-source community core contributors and agent developers to explore commercialization paths, and support a batch of high-growth-potential entrepreneurial newcomers.)

This is the first administrative-document recognition in China that “open-source community core contributor” (开源社区核心贡献者) is a named entrepreneurial subject class with an encouraged commercialization path. Prior briefings documented China’s open-source institutional moves at the state-coordination layer (COPU), municipal-planning layer (Chongqing), and vendor-official layer (Ren Xudong), but none documented the administrative recognition of open-source contributors as an entrepreneurial subject. From an institutional economics standpoint, this state-side-institutional-moves / contributor-as-subject pairing is a structural finding: the open-source contributor’s institutional identity has been formally elevated from a voluntary labor role to an entrepreneurial subject class recognized by administrative document — a structurally new placement that the prior four briefings did not document together.

Move 3 — Item 13 introduces compute-equity and data-equity arrangements (算力入股、数据入股) as supported entrepreneurial-support mechanisms.

The plan’s item 13 reads: “…探索算力入股、数据入股、投孵联动等新型支持方式。” (…explore new forms of support such as compute-equity arrangements (算力入股), data-equity arrangements (数据入股), and investment-incubation linkages (投孵联动).)

This is a structurally new institutional innovation. Prior briefings documented China’s open-source institutional moves at the state-grading, state-capital-consolidation, municipal-planning, and vendor-official layers, but none documented compute and data being treated as equity inputs into SME ventures rather than market-priced factor inputs. From an institutional economics standpoint, this factor-input / equity-input pairing is a structural finding: the plan formally introduces a mechanism design in which compute and data — the two most contested factor inputs in the frontier-AI ecosystem (see Reuters / Jamestown tiered-governance apparatus, September 5 Ren Xudong essay’s platform-neutrality-as-priced-public-good framing) — are being converted from market-priced factors into equity-in-kind contributions. This is a Williamson-L3-style mechanism design that treats non-monetary inputs as co-investment, and it is a structurally new institutional form in the Great Divergence 2.0 framework’s FLOSS pole.

Move 4 — Item 4 also names “one-person companies / super individuals” (一人公司 / 超级个体) as an encouraged entrepreneurial subject.

The plan’s item 4 explicitly names “one-person company” and “super individual” as micro-subjects receiving inclusive support. From an institutional economics standpoint, this collective-open-source-community / individual-entrepreneurial-subject pairing is a structural finding: the plan is placing the individual contributor — the atomic unit of the open-source labor pool — as an SME-form object under administrative recognition. This is institutionally coherent with, but structurally new relative to, Move 2 above: Move 2 names “open-source community core contributor” as a group class; Move 4 names “one-person company / super individual” as a form class. Together they document that the plan treats the individual open-source contributor as both a group subject and an SME form subject.

Move 5 — The evaluation-metrics clause explicitly names open-source project star counts as one of four official performance indicators.

The plan’s closing paragraph reads: “要积极运用人工智能、大数据分析等数字化手段，结合企业领军人才、发明专利、高被引论文、开源项目星标数等多维指标，主动识别高成长潜力企业。” (Actively use digital tools such as AI and big-data analytics, and combine multi-dimensional indicators such as enterprise leading talents, invention patents, highly-cited papers, and open-source project star counts to actively identify high-growth-potential enterprises.)

This is the first administrative-document use in China of the GitHub-style star metric as an official institutional evaluation indicator. Prior briefings documented China’s open-source institutional moves at the state-grading, state-capital-consolidation, municipal-planning, and vendor-official layers, but none documented the GitHub-style star metric being formally adopted as one of four national-level evaluation indicators. From an institutional economics standpoint, this community-norm / administrative-metric pairing is a structural finding: the GitHub-style star count — a community-norm metric that no administrative body previously formally adopted — is now one of four national-level evaluation indicators for identifying high-growth-potential AI SMEs, positioned alongside invention patents and highly-cited papers (both administrative-metric indicators). The institutional-economics reading is that the plan is placing a community-norm metric inside the same evaluation frame as the two conventional administrative-metric indicators, and this placement is the plan’s most institutionally significant act of administrative absorption of open-source community norms.

Why the plan matters for the Great Divergence 2.0 framework

The plan is best read as the sixth institutional surface on which the governance-layer-deficit question has been documented this week. Reading the five briefings of the week together:

  • Xie Lan’s Great Divergence discourse (August 31) — policy-discourse layer
  • COPU’s KubeCon China address (September 2) — national-coordination layer
  • Chongqing’s 15th-Five-Year-Plan (September 2) — municipal-planning layer
  • CAC’s Qinglang Phase 2 (September 3) — administrative-control layer
  • Ren Xudong’s essay (September 4, published September 5) — vendor-official analytical layer
  • MIIT’s AI SME Startup Support Plan (issued August 29, released September 4) — administrative-industrial-policy layer (new)

The plan is the sixth surface — a structurally new placement on the same unsolved object. From an institutional economics standpoint, this five-surface / six-surface pairing is a structural finding: the plan documents that the governance-layer-deficit question has crossed from five documented surfaces into a sixth — and that the sixth surface is the one that treats open-source community norms (star counts) as administrative metrics and open-source contributors as administrative subjects.

One perspective, not a verdict.

The plan is best read as an observation of institutional movement in progress, not as a verdict on institutional direction. AtomGit’s national-community designation does not resolve the atom-git / github-fork question; the contributor-as-entrepreneurial-subject framing does not resolve the compensation-mechanism question; the compute-equity and data-equity arrangements do not yet have implementing guidelines; the star-count evaluation metric does not yet have a national-level calibration standard. What this cycle’s briefing documents is that the governance-layer-deficit question has crossed a first-documented threshold on a sixth institutional surface — and that the sixth surface sits on the same unsolved object as the five documented in the prior four briefings of the week.

Sources

🔍 WeChat Monitor — 16 Official Accounts Tracked

The 16 monitored WeChat official accounts were checked against the September 5, 2026 input file. The MIIT AI SME Startup Support Plan was sourced from the “其他 / Other” section of the input file (via 信通院 CAICT WeChat reprint), not from any of the 16 monitored accounts.

No new institutional content on September 5, 2026 from the 16 monitored accounts:

  1. 开放原子开源基金会 (OpenAtom Foundation)
  2. CCF开源发展技术委员会 (CCF)
  3. 华为开源 (Huawei Open Source)
  4. 木兰开源社区 (Mulan)
  5. COPU 开源联盟
  6. 天工开物开源基金会 (Tiangong Kaiwu)
  7. 明说开源 (Ming Says)
  8. 智源 FlagOpen (BAAI FlagOpen)
  9. 开源社 KAIYUANSHE
  10. 硅基时间 (Silicon Time)
  11. AtomGit
  12. Linux内核开发者大会 (LKDC)
  13. GOSIM / CSDN
  14. 黄大年茶思屋 (Huang Danian Tea House)
  15. 夜天之书 (Yetian’s Book)

Deduplication note: The MIIT AI SME Startup Support Plan is genuinely new institutional content — no prior briefing documented an MIIT-office-level 15-item AI startup support programme at the national level. The plan’s five institutionally-new moves (AtomGit national-community designation, contributor-as-entrepreneurial-subject recognition, compute-equity / data-equity arrangements, one-person-company recognition, star-count evaluation metric) are each first-documented administrative-document moves on the same unsolved institutional object as the prior four briefings of the week.

🔍 Commentary

Six institutional surfaces on the same unsolved object — the governance layer.

This cycle’s single story documents the sixth institutional surface on which the governance-layer-deficit question has been documented in the past week. Prior Chinese-side institutional framings were all state-side, municipal-side, or vendor-side. The MIIT AI SME Startup Support Plan is the sixth — the administrative-industrial-policy layer — and it is the only one of the six that treats open-source community norms as administrative metrics and open-source contributors as administrative subjects.

Five structural findings, all on the same unsolved object.

Reading the plan on the same axis as the prior five briefings of the week documents a coherent six-axis observation on the same unsolved institutional object — the governance layer:

  • Xie Lan’s Great Divergence discourse (August 31) places the open-weight object at the policy-discourse layer — Great Divergence as a term adopted by China’s senior-most AI-governance academic voice.
  • COPU’s KubeCon China address (September 2) places the open-weight object at the national-coordination layer — Lu Shouqun’s 50-vs-2 diagnostic addressed to a Western foundation.
  • Chongqing’s 15th-Five-Year-Plan (September 2) places the open-weight object at the municipal-planning layer — open source as “industrial infrastructure” with compliance codified in the same document.
  • CAC’s Qinglang Phase 2 naming (September 3) places the open-weight object at the administrative-control layer — Doubao / Yuanbao / Qwen / WenxinYiyan named as labeling-compliance subjects.
  • Ren Xudong’s essay (September 4, published September 5) places the open-weight object at the vendor-official analytical layer — the first systematic institutional framework from a Chinese vendor official on the governance-layer-deficit question.
  • MIIT’s AI SME Startup Support Plan (September 4, published September 6) places the open-weight object at the administrative-industrial-policy layer — the first administrative-document moves treating AtomGit as national-community-designated, open-source contributors as entrepreneurial subjects, compute and data as equity inputs, and GitHub-style star counts as official evaluation metrics.

All six framings, on six different institutional surfaces, identify the same unsolved object: the governance layer is where China’s institutional deficit is deepest — a structurally coherent six-axis observation on the same object that prior briefings did not document together.

One structural risk across the six surfaces.

Reading the six surfaces together, the structural risk is that the administrative-industrial-policy surface (the plan) is institutionally the most aggressive absorber of open-source community norms — it is the only surface of the six that formally adopts the GitHub-style star count as an official evaluation metric and the open-source contributor as an official entrepreneurial subject. This absorption is institutionally coherent with the state-grading, municipal-planning, and administrative-control surfaces documented in the prior four briefings, but it is institutionally distinct from the vendor-official analytical surface (Ren Xudong’s essay), which explicitly identifies the governance layer as China’s deepest deficit and calls for Chinese participation in Western neutral foundations (AAIF, MCP, Open VSX) rather than for further Chinese administrative absorption of community norms.

The institutional-economics question — which this cycle’s briefing documents but does not answer — is whether the six surfaces can hold simultaneously without one converting the other. If the administrative-industrial-policy surface (the plan) accelerates, the vendor-official analytical surface’s participation-at-Western-foundations call becomes harder to make from inside the same institutional environment. If the vendor-official analytical surface accelerates, the administrative-industrial-policy surface’s administrative absorption of community norms becomes harder to justify. The plan is the sixth surface on the same unsolved object; it does not resolve the object, it documents the object’s sixth institutional surface.

One perspective, not a verdict.

The MIIT AI SME Startup Support Plan is best read as an observation of institutional movement in progress, not as a verdict on institutional direction. AtomGit’s national-community designation does not resolve the atom-git / github-fork question; the contributor-as-entrepreneurial-subject framing does not resolve the compensation-mechanism question; the compute-equity and data-equity arrangements do not yet have implementing guidelines; the star-count evaluation metric does not yet have a national-level calibration standard. What this cycle’s briefing documents is that the governance-layer-deficit question has crossed a first-documented threshold on a sixth institutional surface — and that this threshold sits on the same unsolved object as the five documented in the prior four briefings of the week.

Editorial note on perspective: This briefing presents one institutional-economics reading of Chinese open-source developments, not a verdict. The “institution” in this story — the administrative-industrial-policy layer, the national-community designation, the contributor-as-subject recognition, the compute-equity / data-equity arrangement, the star-count evaluation metric — is treated as an object of observation, not a target of critique. The Great Divergence 2.0 framework (FLOSS vs. State-Chartered Codebase vs. Intranet Shared Source vs. Cyber-Estate) is a lens, not a universal answer. One perspective, not a verdict.