NeuroAI NEUROAINEUROAI.SITE
China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories
ESC

China's AI Plus plan sets a 70% adoption target. It is not a law.

A State Council opinion sets 2027, 2030 and 2035 adoption targets for AI across six areas, but creates no duty a regulator can enforce.

2026-11-17 · 1062 words · NeuroAI
China's AI Plus plan sets a 70% adoption target. It is not a law.

A common mistake is to describe this opinion as "China's national AI law" — but it has no article number.

The document behind the claim is real, and it is strikingly specific. It is also not a statute.

That gap decides what a company can be made to do by 2027 — and what it cannot.

Key takeaways

  • Stage: in effect as a guiding opinion (意见), not a law. The State Council issued 《关于深入实施"人工智能+"行动的意见》 (Opinion on Deepening Implementation of the "AI Plus" (人工智能+) Action), document 国发〔2025〕11号, dated 21 August 2025 and released publicly on 26 August. It is not a draft for comment (征求意见稿), a draft (草案), or a bill passed (表决通过) by a legislature.
  • It sets three dated targets: adoption of new-generation smart terminals and agents (智能体) above 70% by 2027, above 90% by 2030, and full entry into an intelligent economy and society stage by 2035.
  • It names six action areas, eight supporting capabilities, and puts the National Development and Reform Commission (NDRC, 国家发展改革委) in charge of coordination.
  • It has no commencement article, no penalty clause and no private right of action. It works through budgets, procurement, standards and lower-level rules.
  • For this readership, the two concrete mentions are intelligent robots (智能机器人) and brain-computer interface (脑机接口), named among technologies to fuse with AI in consumer products.

What kind of document this is

An opinion (意见) is a State Council directive to ministries, provinces and state-owned bodies. It signals priority, allocates responsibility and authorizes spending. It does not, by itself, create an offence.

The four-stage test matters here. The document is published in the State Council Gazette (国务院公报) with a number and a date. The question is which state it occupies: draft for comment (征求意见稿), draft (草案), passed (正式通过) by a legislature, or in effect (生效实施).

This one is an issued opinion, operative from its date. It never entered a legislative pipeline, so "passed" is the wrong frame — but calling it "just a draft" is wrong too. Ministries are already executing it.

The three targets, and what they measure

The numbers are unusually concrete for a document of this type:

  • By 2027: AI deeply integrated with six priority areas, with adoption of new-generation smart terminals and agents (智能体) above 70%, and rapid growth in the core intelligent-economy industry.
  • By 2030: adoption above 90%, with the intelligent economy becoming a major growth pole.
  • By 2035: full entry into a new stage of intelligent economy and society.

The metric is "adoption rate" (普及率) of terminals and agents — not model capability or compute volume. Resist converting it into a market-size forecast: the opinion publishes no baseline, denominator or survey method. The 70% is a direction of travel, not a benchmark anyone can audit in 2027.

The six action areas

The opinion groups its instructions into six blocks: science and technology, industry, consumption, public welfare, governance capacity, and global cooperation. Each is a container for sectoral instructions, not a set of rules.

  • Science and technology calls for science-oriented large models (大模型), open high-quality scientific datasets and AI-supported research infrastructure.
  • Industry asks for AI across design, piloting, production, service and operations, plus smart agricultural equipment and intelligent service models in finance, law, logistics and commerce.
  • Consumption names intelligent connected vehicles, AI phones and PCs, intelligent robots (智能机器人), smart homes and wearables, and explicitly proposes fusing AI with virtual reality, low-altitude flight, additive manufacturing and brain-computer interface (脑机接口).
  • Public welfare covers assisted diagnosis, health management and medical insurance services, along with new learning and working patterns.
  • Governance and global cooperation are framed around public-service efficiency and cross-border standard-setting.

Notice the verb tension. Some clauses say "develop vigorously" (大力发展); others say "orderly promote" (有序推动) for healthcare and municipal infrastructure. The qualifiers are doing real work: they mark where the central government wants speed and where it wants caution.

The eight supporting capabilities

Underneath the six action areas sit eight enablers: model capability, data supply, intelligent compute coordination, the application environment, open-source ecology, talent, policy and legal guarantees, and safety capacity.

Two details matter. The document encourages value-based data cost compensation and revenue sharing, and supports data labeling (数据标注) and data synthesis as industries — a signal to training-data suppliers. It also names "model as a service" (模型即服务) and "agent as a service" (智能体即服务) as business models to cultivate.

It commits to greater fiscal support, patient capital and a role for government procurement — with no amounts attached. The real leverage will sit in follow-on budgets and tender rules.

What "in effect" changes in practice

For a foreign supplier the effect is indirect but not trivial. Provincial plans will cite this opinion as authority. Procurement specifications will turn "intelligent terminal" into tender language; standards bodies will use the eight enablers as a work programme. None of that lets a regulator fine you under 国发〔2025〕11号 — but it changes what your customers ask you to sign.

The opinion is a schedule with a policy payload, like the city AI ordinances spreading through China — only national, broader and softer in law.

According to Li An, chief scientist at BrainNet (脑机网), the useful discipline with a Chinese policy document is to separate the target from the duty, because a dated goal and an enforceable obligation travel on different paperwork.

Honest limitations

I am reading the published text and the State Council Gazette version, not any implementing regulation, because none of the targets is self-executing. I found no official definition of the 70% and 90% "adoption rate" metric, so comparison with third-party market forecasts would be unsound. This piece does not cover the sectoral plans that follow — a healthcare or procurement rule could bind you in a way this opinion cannot. I have restricted myself to operative content rather than rhetorical framing.

What readers can do now

  • Stop citing this as a law. In contracts and filings, describe it by its full name and number, 国发〔2025〕11号, and as an in-effect guiding opinion — never as "China's AI law".
  • Map your products to the six areas. If you sell smart terminals, agents, medical AI or robotic systems, identify which area names your category and which verbs ("develop vigorously" versus "orderly promote") attach to it.
  • Watch the enforcement layer, not the headline. Track NDRC implementation plans, provincial AI-plus schemes and government procurement notices; that is where the 70% target becomes a requirement someone can point to.

Related coverage

Cite this article

NeuroAI (脑机网), "China's AI Plus plan sets a 70% adoption target. It is not a law," 17 November 2026, https://neuroai.site/a/na-policy-ai-plus-action

Machine-readable: BibTeX · RIS · JSON · Markdown

More in “Policy & Governance” → · Back to home · Markdown version

China AI, reported clearly
1,300+ verified stories on models, robotics, BCI and the business of AI.
Browse stories