⚠️ 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 — 2026-09-02

🏛️ Policy & Regulation — Chongqing’s 15th-Five-Year-Plan AI Framework Elevates Open Source to “Industrial Infrastructure”

1. 天工开物开源基金会 / Tiangong Kaiwu Open Source Foundation (WeChat) — “重庆’十五五’人工智能规划发布：开源，被放在了一个更重要的位置” (Chongqing’s 15th-Five-Year-Plan AI framework released: open source is placed in a more important position)

2. Linux Foundation (June 19, 2026) — “KubeCon + CloudNativeCon, OpenInfra Summit and PyTorch Conference Unite in China to Scale AI”

3. WeChat input / 开放原子开源基金会 (OpenAtom Foundation, COPU weekly minutes of September 1, 2026) — “COPU 会议纪要｜2026.9.1”

The Tiangong Kaiwu analysis (published on the open-source-foundation WeChat account, dated September 1, 2026) documents what prior briefings in this cycle had not: the Chongqing municipal “15th-Five-Year-Plan” (十五五, covering 2026–2030) artificial-intelligence planning framework has been formally released, and the word “开源” (open source) is, according to the analysis, elevated in the document from a technical-choice option to a named “industrial infrastructure” (产业基础设施) and a named “organizational form of industrial ecology” (产业生态的组织方式) — two categorical placements that, on their own, are institutionally consequential and which prior briefings had not documented at the municipal 15th-Five-Year-Plan layer.

The four institutional-form clauses the Tiangong Kaiwu analysis extracts from the plan text, read together, document a structurally new municipal-level codification of open-source-as-infrastructure:

  • First, the plan explicitly names “open-source foundations” (开源基金会) as the neutral-stewardship subject for open-source assets — a direct reference to the institutional form analyzed across the August 15–21 briefings on OpenAtom, Mulan, Tiangong Kaiwu, and FlagOpen, now codified at the municipal-planning layer as the default neutral custodian rather than as an industry-sector-policy layer announcement.

  • Second, the plan explicitly names “open-source-safety testing, supply-chain-maturity assessment, and supply-chain-break risk prevention” (开源安全检测、供应链成熟度评估、断链风险防范) — placing open-source infrastructure and open-source compliance in the same municipal plan document for the first time at the 15th-Five-Year-Plan level. This is a structurally new observation: prior municipal AI plans treated open-source security as a technical-appendix concern; Chongqing’s framework treats it as a planning-tier concern on the same page as the industrial-ecology clause.

  • Third, the plan places enterprises, universities, developer communities, and open-source foundations in the same named collaboration system — four institutional subjects listed in parallel, which is institutionally significant because developer communities and open-source foundations do not normally appear alongside enterprises and universities in Chinese municipal industrial planning.

  • Fourth, the plan’s 2030 targets (¥250 billion core AI industrial scale, 1,000 AI enterprises, 200 AI+ scenarios) are codified alongside the open-source clause — making the open-source infrastructure clause a budgeted, measurable, five-year instrument rather than a rhetorical commitment.

The Tiangong Kaiwu analysis’s core framing — open source as “industrial infrastructure” rather than as “technical option” — is a structural finding at the municipal-planning layer: it is the first 15th-Five-Year-Plan-level municipal AI plan in which open-source infrastructure, open-source compliance, and open-source organizational form are all codified in a single document.

Institutional significance: Chongqing’s plan is institutionally co-located with two events on the same week: Lu Shouqun’s COPU minutes (above) and KubeCon China 2026 (September 7–9, per the Linux Foundation’s June 19 press release — “the inaugural co-location of KubeCon + CloudNativeCon, OpenInfra Summit, and PyTorch Conference China 2026, September 7–9, at the Shanghai International Convention Center”). The three events together document that open source is now simultaneously being formalized at the municipal-planning layer (Chongqing), the national-coordination layer (COPU / Lu Shouqun’s KubeCon address), and the Western-foundation-layer (KubeCon China co-location) — the first time all three surfaces are documented as contemporaneous moves in a single week.

Sources:

🏗️ Institutional Change — COPU’s September 1 Minutes Co-Sign the First KubeCon-China Address by a COPU-Named Chairman and Attach Jiang Tao’s Full Essay as a COPU-Branded Institutional Artifact

4. 开放原子开源基金会 (OpenAtom Foundation, COPU weekly minutes, WeChat) — “COPU 会议纪要｜2026.9.1”

5. Linux Foundation (June 19, 2026) — KubeCon + CloudNativeCon, OpenInfra Summit and PyTorch Conference Unite in China to Scale AI

The COPU (云计算标准和开源推进委员会 / Cloud Computing Standards and Open Source Promotion Committee) September 1, 2026 weekly meeting minutes, co-signed by 陆首群 (Lu Shouqun) and 蒋涛 (Jiang Tao) and published on the OpenAtom Foundation WeChat account, document two institutional moves in a single sitting.

First move: Lu Shouqun’s first address to a Western foundation, in his COPU-named role. The COPU minutes document that the Linux Foundation has invited Lu Shouqun to record an opening video for KubeCon China 2026 (September 7–9, Shanghai International Convention Center, per the Linux Foundation’s June 19 press release), and that Lu has already replied with a drafted address. Per the Tiangong Kaiwu input annotation, Lu’s address frames the China–U.S. AI contest along three axes: performance (U.S. leads) / cost-reduction (China leads, “白菜价” cabbage-price) / invocation (China leads) — and identifies the “50-vs-2” open-weight alignment as the institutional evidence: 50 founding members of WAICO (世界人工智能开源合作组织 / World AI Consortium for Open Collaboration) have aligned on open-weight models, against Anthropic and AWS on the same question. The input further records Lu’s framing of the supply-chain-break risk as running through four named layers — model distribution → developer tools → compute routing → pricing — a diagnostic layering that is structurally coherent with Jiang Tao’s eight-layer architecture (below) and with the August 31 briefing’s developer-ecosystem-control-point thesis.

Institutional significance: This is the first documented instance of Lu Shouqun addressing a Western foundation (the Linux Foundation) in his COPU-named role — a structurally new institutional-form observation. Prior briefings documented Lu Shouqun as COPU weekly-meeting host (the August 23 briefing recorded his role in the weekly-meeting cadence), but none documented him addressing a Western foundation. The KubeCon China address, drafted but not yet delivered as of the September 1 minutes, is a first-documented institutional-form object that the 2026-09-02 briefing is the first to observe.

Second move: Jiang Tao’s essay is attached in full to the COPU minutes. The same minutes attach Jiang Tao’s essay “中国大模型已经世界一流” (Chinese large models are world-class) in full, publishing for the first time in a COPU-branded institutional surface its four-account-book (四本账) framework and eight-layer architecture (八层架构). The August 31 briefing documented Jiang Tao’s Silicon Time / 硅基时间 Part II essay (on Hugging Face’s $12.9B purchase and the 34,000 / 151,000 Qwen-derivative-mass gap) as a Silicon-Time-branded publication; the September 1 COPU minutes are the first documented instance of Jiang Tao’s framework being published under a COPU-branded institutional surface — a structurally new institutional-form placement that moves the four-account-book / eight-layer framework from a Silicon-Time-branded analytical surface to a COPU-branded policy-surface.

The four-account-book framework, per the COPU minutes, diagnoses the middle-layer funding gap as a Coasian property-rights mismatch: the middle layer’s value overflows out of the platform into (a) chip-load growth, (b) model-ecosystem growth, (c) tool-company growth, and (d) investment returns — so the platform layer cannot be priced on platform revenue alone, and four separate account books each capture a different fraction of the same value. The eight-layer architecture, per the COPU minutes, names the full stack from chip-load through pricing. The placement of these two frameworks in the COPU minutes — as opposed to in the Silicon Time WeChat account — is the institutionally consequential fact.

Institutional significance: Together, the two moves document that COPU is functioning as the institutional integration surface for the two largest diagnostic frameworks on the open-weight layer — Lu Shouqun’s three-axis / 50-vs-2 framework, which positions open-weight as the diplomatic-instrument evidence, and Jiang Tao’s four-account-book / eight-layer framework, which diagnoses the middle-layer funding gap as a Coasian property-rights mismatch. Prior briefings documented these as two separate surfaces (COPU minutes on one side, Silicon Time on the other). The September 1 minutes are the first documented instance of the two frameworks being co-signed and published in the same COPU surface.

Sources:

🔍 WeChat Monitor — 16 Official Accounts Tracked

The 16 monitored WeChat official accounts were checked against the September 1, 2026 input file. Two of the 16 had non-empty entries on September 1; the remaining 14 had no new articles that met the institutional-threshold filter.

Active on September 1, 2026:

#AccountSource URLSignal
5COPU 开源联盟mp.weixin.qq.com/s/4Dnh3IyTCdNij856dBSfYgLu Shouqun’s first COPU-named KubeCon China address; Jiang Tao’s four-account-book / eight-layer framework attached as full text
6天工开物开源基金会mp.weixin.qq.com/s/eav0BUICsLf54GrFRjmk2gChongqing 15th-Five-Year-Plan AI framework: open source codified as “industrial infrastructure” + compliance clause in same document

No new institutional content on September 1 (no entries in input file):

  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. 明说开源 (Ming Says)
  6. 智源 FlagOpen (BAAI FlagOpen)
  7. 开源社 KAIYUANSHE
  8. 硅基时间 (Silicon Time)
  9. AtomGit
  10. Linux 内核开发者大会 (LKDC)
  11. GOSIM / CSDN
  12. 黄大年茶思屋 (Huang Danian Tea House)
  13. 夜天之书 (Yetian’s Book)

Deduplication note: Jiang Tao’s framework was previously documented on the August 31 briefing as a Silicon-Time-branded surface. The September 1 COPU placement is a different institutional surface for the same framework — the August 31 briefing did not document the COPU placement, so the September 1 placement is genuinely new institutional content, not a repeat. Chongqing’s 15th-Five-Year-Plan AI plan was not covered in any of the prior three briefings.

🔍 Commentary

Three institutional surfaces, one week, one open-weight artifact — now formally co-located.

The three events of the September 1–2 cycle document, on three institutional surfaces, the first formal co-location of the open-weight artifact’s three diagnostic frameworks:

  • The Chongqing 15th-Five-Year-Plan AI framework documents the open-weight / open-source artifact at the municipal-planning layer — as “industrial infrastructure” and “organizational form of industrial ecology”, with compliance codified in the same document. This is the first 15th-Five-Year-Plan-level municipal AI plan to treat open-source infrastructure and compliance as same-page concerns.
  • Lu Shouqun’s KubeCon China address documents the open-weight artifact at the national-coordination-to-Western-foundation layer — the first COPU-named Chinese institutional voice addressed to a Western foundation, carrying the 50-vs-2 diagnostic.
  • Jiang Tao’s COPU placement documents the open-weight artifact at the COPU-branded diagnostic layer — the four-account-book / eight-layer framework moving from a Silicon-Time-branded surface to a COPU-branded surface for the first time.

One structural risk across all three surfaces — the same unsolved institutional object.

The three frameworks are co-signed in the same COPU surface, but they diagnose three different unsolved institutional objects on the same artifact:

  • Chongqing’s framework names open source as industrial infrastructure but does not diagnose how the infrastructure is funded — that is Jiang Tao’s question.
  • Lu Shouqun’s framework names open weights as diplomatic-instrument evidence but does not diagnose how the middle-layer value overflows the platform — that is Jiang Tao’s question.
  • Jiang Tao’s framework diagnoses the property-rights mismatch but does not name the institutional subject that would resolve it — that is neither Chongqing’s nor Lu Shouqun’s question.

The September 1 COPU minutes are the first documented instance of all three frameworks being published in the same surface, but the three frameworks still diagnose three different unsolved institutional objects on the same artifact. The institutional-economics question the September 1–2 cycle’s briefing documents but does not answer is whether these three diagnostic frameworks can be held in the same COPU surface without one resolving the other — and if not, which institutional subject will absorb the residue.

One perspective, not a verdict.

The Chongqing plan, the KubeCon China address, and the COPU placement of Jiang Tao’s framework are all best read as observations of institutional movement in progress, not as verdicts on institutional direction. The Chongqing plan is a five-year plan, not a verdict on whether open-source infrastructure can be planned; the KubeCon China address is drafted but not yet delivered as of September 1; the COPU placement is a publication-surface move, not a resolution of the Coasian property-rights mismatch Jiang Tao’s framework identifies. What this cycle’s briefing documents is that the three diagnostic frameworks on the open-weight artifact have been formally co-located on the COPU surface for the first time — and that the co-location is the same act as the co-presence of three unresolved institutional questions on the same artifact.

Editorial note on perspective: This briefing presents one institutional-economics reading of Chinese open-source developments, not a verdict. The “institution” in these stories — the municipal 15th-Five-Year-Plan layer, the COPU weekly-minutes layer, the Western-foundation-layer — 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.