Home › Research › Open-source policy at EU, national, Chinese central and provincial level
Open-source policy at EU, national, Chinese central and provincial level
Draft v0.1, 2 November 2026. Data: ../data/initiatives.json (50 records). Source URLs: ../data/new_source_urls.txt. Method: about 40 web calls. Chinese municipal texts for Beijing, Shanghai, Hangzhou, Shenzhen and Chongqing were read in full or in part through the r.jina.ai relay; MIIT only as a press-conference text. Many EU items rest on search summaries. Each record carries an evidence grade (primary / secondary / unverified). Targets are announced goals, not results, unless stated.
(a) Overview
| Level | Records | Primary | What it contains |
|---|---|---|---|
| EU | 12 | 4 | Open Source Strategy (3 Jun 2026), Maintenance Instrument, Open Internet Stack, Commission OSPO and network, DC-EDIC, EDCOM, EU-STF pilot, Interoperable Europe Act/OSOR, NGI, CRA Art. 24, Digital Europe (not researched), JRC (not researched) |
| CN-central | 11 | 6 | MIIT AI+Software plan (11 Sep 2026), 14th and 15th FYP, OpenAtom, AtomGit, Gitee, Mulan, GB/T 44272-2024 (unverified), KAIYUANSHE, OpenHarmony, RISC-V guidance (unverified) |
| EU-national | 15 | 3 | FR La Suite (+ unsourced Etalab stub), DE ZenDiS/openDesk, Sovereign Tech Agency, Deutschland-Stack, Schleswig-Holstein, IT CAD arts. 68-69 + Developers Italia, NL policy and code.overheid.nl, ES CTT, DK, EE X-Road, AT Bundesheer, BE, PL |
| CN-provincial | 12 | 7 | Beijing, Shanghai, Zhejiang, Hangzhou, Jiangsu/Suzhou, Guangdong, Shenzhen (+Longgang), Chongqing (plan + foundation), Wuhan, Hunan |
Changes to the memo found while verifying:
- MIIT says "5 or more" key projects by 2028, not "at least five". The text read does not mention SBOM, public-sector adoption or international cooperation.
- Beijing's plan was issued 24 Dec 2025. It contains no amounts, but it has a procurement-linked instrument ("open-source first solution", support as a share of actual procurement value) and SBOMs "in principle" for fiscally funded software. The memo underplayed this.
- Shanghai: doc no. 沪府办〔2025〕33号, effective 15 Dec 2025 to 14 Dec 2028. No monetary amounts and no procurement or OSPO provisions.
- Hangzhou caps verified: RMB 1m / 500k, plus RMB 2m platform grants. Hangzhou 杭政办函〔2026〕9号, effective 27 May 2026 to 2028.
- Shenzhen: 深府办规〔2026〕1号, 15% + 5%, RMB 10m cap confirmed.
- Jiangsu: the RMB 1bn is "planned total scale"; paid-in status unstated. 16 first tenants.
- Guangdong: no document number; announced at a press conference (year not on page).
- Chongqing: milestones of 20,000 developers by 2027 (new); no amounts.
- New: Wuhan (Jan 2026), explicitly encouraging fiscally funded public-body systems on OpenHarmony (secondary source); Hunan (13 Oct 2026).
- Not found: Sichuan, Fujian, Shandong provincial open-source plans (one search; absence not proven).
- The 15th FYP Outline (13 Mar 2026): no confirmation of an explicit open-source passage. Only the Government Work Report line "support building AI open-source communities" is reported, via secondary press.
(b) Cross-level comparison
EU vs China central
| Dimension | EU | China central |
|---|---|---|
| Core instrument | Soft law plus institutions: Strategy, DC-EDIC, EDCOM, OSPO network. Hard law at the edges: Interoperable Europe Act (equivalence-based open-source preference), CRA steward role | Industrial-policy plans (MIIT) that channel projects to one national foundation and one national platform |
| Who funds | Commission grants (NGI, EDCOM, possibly Digital Europe), Member State contributions; Open Internet Stack and the Maintenance Instrument have no amounts yet | No central amounts found in the MIIT text; funding is implicit through OpenAtom and provinces |
| Reuse logic | Reuse of public code and interoperability solutions (portal, OSPO network, procurement guidelines, EUPL) | Reuse of code, models and datasets on AtomGit; reuse as hosting/discovery, not as administrative duty |
| Creation vs maintenance | First explicit maintenance tools: Maintenance Instrument, EU-STF pilot, dependency mapping, CRA stewards | Creation and incubation dominate (key projects, incubation, projects count). Maintenance is not a named instrument in the texts read |
| Governance | Multi-state consortium, subsidiarity, neutrality criteria | One national foundation (OpenAtom) acts as steward and platform owner; independence not assessed |
| Public-sector demand | Procurement guidelines and open-source-first language (weak per CNLL) | Not addressed in the MIIT text read; left to provinces |
EU national vs China provincial
| Dimension | EU member states | Chinese cities/provinces |
|---|---|---|
| Typical instrument | Platform or institution run by the administration (La Suite, openDesk, code.overheid.nl, CTT), legal duty to publish (IT art. 69, NL "open unless") | Quantified ecosystem targets plus subsidy: project awards (Hangzhou), community operating-cost share (Guangdong), adoption subsidy (Shenzhen), funds (Jiangsu) |
| Who funds | Central government budgets; DINUM seeks ministry contributions; ZenDiS federal; STA federal | City/province budgets; Jiangsu mixes public money and investor funds (announced) |
| Reuse logic | Reuse between administrations (same code deployed in many bodies; Docs co-developed FR/DE) | Reuse mostly between firms and communities (national hosting required for Hangzhou award); public-sector reuse appears only in Beijing (SBOM, donation to platforms) and Wuhan/Longgang/Chongqing (OpenHarmony systems) |
| Demand side | Sovereignty-driven migration (FR Visio 2027, SH LibreOffice, DK, AT) | Industrial upgrade subsidies (Shenzhen), public-body IoT systems (Wuhan, Chongqing) |
| Maintenance | STA (DE) is the only mature example; most MS programmes build and migrate | Guangdong cost share is the closest analogue; no maintainer-level fund seen |
| Transparency | Usage figures often press/vendor claims, critics exist (Cour des comptes) | Targets only; achieved results rarely reported |
A recurring pattern: Europe's strength is administration-to-administration reuse and law; China's is fast, quantified, multi-instrument local programmes. Neither side evidences independent outcome data.
(c) Pairings
| EU / EU-national | China | Fit | Why |
|---|---|---|---|
| EU-01 Strategy | CNC-01 MIIT AI+Software | partial | Both make open source industrial/sovereignty infrastructure. EU stresses public administration and maintenance; MIIT stresses incubation and AI |
| EU-05 DC-EDIC | CNC-04 OpenAtom | partial | Stewardship institutions; very different mandate (consortium of states vs. non-profit foundation) |
| EU-12 EU-STF, EUN-DE-02 STA | CNP-HZ, CNP-GD, CNP-JS | weak | Money for open-source work. Hangzhou pays one-off incubation awards, Guangdong pays community operating costs, Jiangsu is venture finance. None funds maintainers of third-party dependencies |
| EUN-NL-01, EUN-IT-01 | CNP-BJ | partial | Only Beijing couples procurement and SBOM with open source in a public text |
| EU-09 CRA / E6 dependency intelligence | CNP-BJ, CNP-SH (SBOM) | partial | Both lean on SBOMs; legal force differs (EU regulation vs municipal plan) |
| EUN-FR-01, EUN-DE-01 workspace suites | CNP-SZ-LG, CNP-WH, CNP-CQ (OpenHarmony in government systems) | weak | Same goal (government systems on open base) but an OS/IoT base is not a document and messaging suite |
| EU-04 OSPO network | CNP-BJ open-source promotion centres | weak | Functional similarity (advice, assessment), no network |
Pairing is impossible for: the Interoperable Europe Act (no Chinese interoperability-sharing law found), CAD art. 69 and ENI-style mandatory publication (the earlier OFE-based finding of no formal public-administration reuse policy still holds at national level), the Collaborative Workspace flagship (E5, no equivalent, as in the memo), NGI-type micro-grants (nothing comparable found), CRA stewards (no Chinese legal category found).
(d) Bridge opportunities at operational level
All are research or evidence exchanges. None implies funding or endorsement, and each is subject to the constraints in memo §8.
| # | Bridge | Level | Actors | Why feasible |
|---|---|---|---|---|
| 1 | Procurement and SBOM comparison note | EU-national / CN-provincial | Beijing EITB, NL BZK OSPO, Italy AgID | Beijing's text is public and quantifiable; NL and IT have written duties. A desk comparison needs no agreement |
| 2 | Maintenance-funding design comparison | EU / CN-provincial | EU-STF pilot, Sovereign Tech Agency, Hangzhou, Guangdong, OpenAtom | Both sides publish caps/rates; compare grant, cost-share and venture designs and what each leaves unfunded |
| 3 | Public-body OpenHarmony vs open-suite deployments: lessons on adoption barriers | EU-national / CN-provincial | DINUM, ZenDiS, Longgang, Wuhan | Shared problem (moving administrations off proprietary stacks); Cour des comptes critique and Longgang pilot give contrasting evidence. Not a technical exchange |
| 4 | Shared vocabulary for "digital commons" and ownership/non-capture rules | EU / CN-provincial | DC-EDIC (E3), Shanghai Open Source Association, KAIYUANSHE | Shanghai's association already uses digital-commons language; start from the declaration text once obtained |
| 5 | Common reading of state-of-open-source reports | EU / CN-central | E3 State of Digital Commons, KAIYUANSHE China Open Source Report, JRC, OSOR | Both publish; compare metrics and definitions to avoid reading Chinese figures as European-style indicators |
| 6 | Licence and compliance mapping | EU / CN-central | Interoperable Europe/EUPL, OpenAtom, Mulan, GB/T 44272 | Mulan PSL v2 is OSI-approved; EUPL compatibility analysis is a bounded legal task. First verify GB/T 44272 |
| 7 | Eclipse-OpenAtom channel as existing neutral venue | EU / CN-central | Eclipse Foundation, OpenAtom, Oniro | Agreements and sister-project mechanism already exist (S25-S28) |
| 8 | Critical-dependency and project-health metrics | EU / CN-central | E6 dependency intelligence, JRC, OpenAtom/AtomGit, X-lab | Open-data metrics are public; method comparison avoids sharing sensitive data |
| 9 | Conference-level exchange on project-count targets vs outcomes | EU-national / CN-provincial | Italy Developers Italia, Zhejiang, Shanghai | Italy's catalogue gives measured counts, Shanghai/Zhejiang give targets; compare target design and measurement |
| 10 | Education and contribution-credit practice | EU-national / CN-provincial | Austria/Bundesheer upstream contributions, Zhejiang evaluation rules, Chongqing curricula | Zhejiang credits contributions in appraisal; European public bodies could test equivalent practice |
| 11 | X-Road and similar cross-border platforms vs Chinese multi-jurisdiction platforms | EU-national / CN-provincial | NIIS, Shanghai/Hangzhou platforms | Governance of jointly run open platforms; weak evidence on the Chinese side, so scoping first |
| 12 | Open Internet Stack vs AtomGit catalogue design | EU / CN-central | Commission, OpenAtom | Both build a catalogue/hosting layer; compare openness, mirroring and exit rules (mirroring is in the EU Strategy) |
Priority for next work: 1, 2, 4. They rely on published texts, need no data sharing, and match E2, E6/E7 and E3.
(e) Gaps and unknowns
- Not verified or not found: 15th FYP Outline open-source text; GB/T 44272-2024 (title seen only in a listing); CAC and other ministries (nothing found); national RISC-V guidance; Sichuan, Fujian, Shandong; Jiangsu fund paid-in status; Shanghai declaration text; Etalab/French legal base; German OZG; Spain after 2025; Denmark 2026; Estonia, Poland, Belgium country reports (summary only); Digital Europe and JRC (not researched); art. 68 CAD text.
- EU budgets: no amount for the Maintenance Instrument or Open Internet Stack; the EUR 2bn / seven years figure is secondary. EDCOM's size and funder are unknown. EU-STF start date conflicts (March vs June 2026).
- Chinese texts read through secondary reports: Zhejiang, Jiangsu, Guangdong, Wuhan, Hunan. Official texts would be needed for quotation.
- Outcome evidence: neither side offers independent results. Chinese items are almost all targets; French and German adoption figures come from the operators.
- Structural: Chinese public-administration reuse remains an open question; Wuhan, Longgang and Chongqing suggest OpenHarmony pilots rather than a reuse regime. Revisit the OFE finding that there is no formal policy.
- Cautions (memo §8): export controls, data transfer, dual-use, political sensitivity. Nothing here is an agreed position of the DC-EDIC.
Round 2 findings (2026-11-02)
Light pass (about 15 searches/fetches). Catalogue now 56 records (added CNC-12..15, CNP-HB-WH, EU-13); notes appended to EU-01, EU-05, CNP-BJ/ZJ/HN.
- Shanghai September 2026 declaration (priority 1): not found. Chinese and English searches return only the Dec 2025 Shanghai open-source plan and the 12 Sep 2026 Pudong Innovation Forum. Treat the declaration, including its title, as unconfirmed until a primary text appears. Listed in data/cn_vpn_needed.txt.
- 15th FYP Outline: Xinhua published the full text on 13 Mar 2026 (18 parts incl. Digital China); the fetched page held only the part list, so no 开源 passage could be confirmed or ruled out. Still open.
- MIIT: 4 Sep 2026 AI SME start-up plan 2026-28 (AtomGit, funds for open-source projects); 13 Jan 2026 industrial-internet platform plan (open-source ecosystem goal); on 20 Jul 2026 MIIT said an "AI+software" action plan is coming and 50+ projects are incubating at the national foundation. No joint MIIT/NDRC open-source document found.
- GB/T 44272-2024: verified on the SAMR openstd listing (current; published 2024-08-23; effective 2025-03-01). Full text not read.
- SASAC (new): Huanxin AI open-source community v2.0, July 2026, China Mobile-led; Feb 2026 directive to central SOEs on open-source communities. Figures are press-only. CAC, MOST, MOE: nothing found.
- Public-administration reuse: only secondary English reports on replacing government Windows with Kylin/UOS (openEuler/openKylin as upstreams); no official source, nothing on code sharing between agencies. The no-formal-reuse-regime finding stands.
- Provinces: Wuhan (4 OpenHarmony OS, 100k developers by 2027) and Hunan (doc no. 湘制造强省办〔2026〕7号, 13 Oct 2026: 100 scenes, 50k developers by 2028) now have dated targets, still secondary. Sichuan/Chengdu, Fujian, Shandong: nothing found.
- EU: strategy is COM(2026) 503. No official budget for the Maintenance Instrument or Open Internet Stack; EUR 350m for an EU-STF is a study/advocacy figure. DC-EDIC (FR, DE, NL, IT, LU; Paris; director from 1 Apr 2026) has no published budget; a Digital Europe EDIC Support Hub call exists (EUR 1m ceiling per an aggregator, to check). CRA: non-binding guidance C(2026) 5252 of 27 Jul 2026 covers FOSS; no separate steward act; reporting duties from 11 Sep 2026, main duties 11 Dec 2027. EU OSPO and EU-STF start date: not researched.
- Catalogue upgrades (priority 8): not done beyond the above, for budget reasons.
Round 4 counts (2026-11-02)
Registry now 62 records, 41 primary (66%). Upgrades and additions are listed in docs/CATALOGUE-VERIFICATION-NOTES.md (Round 4).
| Level | Primary | Secondary | Unverified | Total |
|---|---|---|---|---|
| EU | 10 | 2 | 1 | 13 |
| China central | 7 | 5 | 3 | 15 |
| EU member states | 13 | 3 | 2 | 18 |
| Chinese provinces / cities | 11 | 5 | 0 | 16 |
Guangdong/Guangzhou finding: provincial and municipal bodies have shipped HarmonyOS versions of government apps (Yue Zhengyi Aug 2025, Sui Hao Ban Sep 2025; facts) and Guangzhou commits to converting government terminals by 2028 (commitment). The sources say HarmonyOS, not OpenHarmony; the open-source link is an inference.