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

China Issued 18 Measures to Make AI Adoption Friendlier to Jobs

A guiding opinion published on October 23, 2026 sets out 18 measures on AI and employment — from new occupations and retraining to job-impact reviews and unemployment buffers — and it is already in force as policy.

2026-11-19 · 1168 words · NeuroAI
China Issued 18 Measures to Make AI Adoption Friendlier to Jobs

A factory technician watches a new inspection system flag a defect in half a second. The system is faster than she is. Whether she still has a job in two years is now a question Chinese policy is trying to answer in advance.

On October 23, 2026, the Chinese government published a guiding opinion on adapting to artificial intelligence in order to promote employment-friendly development. It is an official attempt to pair AI adoption with job protection rather than treating the two as a trade-off.

Key takeaways

  • The document is a guiding opinion (意见) issued by the General Office of the State Council, numbered Guoban Fa [2026] No. 29, dated October 16 and published October 23, 2026.
  • It sets out 18 measures across six policy areas — the official text has eight sections in total, including general requirements and implementation.
  • Measures include developing new AI-related occupations, requiring employment-risk assessments for certain market-access projects, encouraging an unemployment buffer period (失业缓冲期) and expanding retraining.
  • It is a formally issued and published policy document, not a draft — but it is not legislation passed by the national legislature either. Those are different stages, and the difference matters.

First, place it in the right stage

Chinese policy documents move through four recognizable stages, and blur them at your peril.

A draft or consultation document (征求意见稿/草案) proposes a rule and asks for comment. A formally adopted or issued document (正式通过/发布) has crossed that line. An effective document (生效实施) can be applied. Finally comes implementation (贯彻落实), where ministries and local governments write the supporting rules and start enforcement.

The October opinion sits at stage three moving into stage four. It was signed on October 16, published on October 23, and took effect as a policy directive on publication. It was never a draft in this form, and it is not a law passed by the National People's Congress. A guiding opinion sets direction for government bodies; the concrete effect for a worker or an employer will depend on the implementing rules that follow.

"Issued" is not "drafted," and "guiding opinion" is not "statute."

What the 18 measures actually say

The six substantive areas are easy to summarise in plain English.

Create jobs. Deepen the "AI plus" (人工智能+) action, support small and micro enterprises, use government funds such as the national AI industry investment fund to back AI businesses, and build open-source communities that spin off new companies. The document also calls for developing new occupations (新职业) and occupational standards, naming AI engineers, AI trainers and digital-management specialists.

Make adoption less disruptive. Prioritise AI in jobs with labour shortages or high safety risk, and be "steady" about the scope, pace and intensity of AI use in labour-intensive industries. Employers are encouraged to explore an unemployment buffer period: giving affected workers advance notice and transition support before a large-scale rollout.

Assess the risk before approval. For AI application projects seeking market access in fields where the employment impact could be large, a job-impact assessment should be carried out during review. Other fields should conduct assessments where relevant. This is probably the most operationally significant clause in the document, because it attaches a review step to a decision point.

Train people. Expand AI-related training through universities, vocational schools and public platforms, including short, low-cost courses for graduates, migrant workers and people who have lost jobs. General AI literacy is to be promoted through public education platforms.

Upgrade employment services. Use AI to match workers to policies and jobs, improve national employment platforms, and support new products such as job-hunting and recruitment large models (大模型), intelligent interviews and virtual-reality training.

Protect workers and monitor risk. Study rights in human-machine collaboration, curb algorithmic discrimination and false recruitment in hiring, extend social insurance to flexible and new forms of employment, and fold AI-driven job losses into existing mass-unemployment response mechanisms. A monitoring and early-warning system is to be built.

The numbers behind the policy

The official figures released alongside the opinion give it scale. China's core AI industry was valued at more than RMB 1.2 trillion (≈ US$169B / HK$1.32T) in 2025, with over 6,200 AI companies. By the end of that year, more than 30 per cent of manufacturing enterprises with annual main business revenue of at least RMB 20 million (≈ US$2.8M / HK$22.0M) had adopted AI technologies.

Demand is moving too: recruitment platform Maimai reported that new AI-related job postings rose about eightfold year on year from January to July. That is one platform's data, not a national census, but it explains the emphasis on skills.

What it does not do

It does not ban AI deployments, set a quota for job cuts, or create a new workplace licence. It does not automatically entitle a displaced worker to a benefit — support clauses generally say "as prescribed," leaving eligibility to other rules. Nor does a guiding opinion change an employment contract by itself.

The honest reading: this is a coordination document that tells agencies what to weigh and gives local governments a framework to fill in. Whether the buffer becomes real practice depends on the measures that follow.

Honest limitations

This article is based on the full official text on the Chinese government website and an English summary released by Xinhua and reprinted by Chinese embassies. I have relied on the original Chinese for the document number, dates, section counts and specific clauses; where the English summary says "18 tasks in six areas," I have kept that phrasing while noting the text itself has eight sections.

I have seen no implementing rules yet, so the practical effect cannot be assessed. The industry and adoption figures are official estimates, not audited statistics, and the Maimai number is one platform's data. I have not interviewed affected workers or employers, and no statement is quoted beyond the published text.

The document's ideological preamble is not reproduced here; this article covers only its employment measures. Chinese terms are given with originals where wording is specific, and translations vary.

What readers can do now

  • Check the stage before you react. If you hear that "China has new AI rules," ask whether you are looking at a draft, an issued opinion like this one, or a binding implementing rule. The obligations differ sharply.
  • If you work in an affected sector, watch for local implementation. Job-impact reviews and buffer periods only become real when a province or ministry publishes the procedure. That is the document to read next.
  • If you are planning a career, take the skills signal literally. New occupational standards and short training programmes are the parts of this policy most likely to reach individuals first.

The technician in the factory may still lose tasks to the inspection system. What this opinion changes is that someone is now required to count the cost before the rollout, and to plan for what comes after.


Sources: State Council General Office, Guoban Fa [2026] No. 29, gov.cn, issued Oct 16 / published Oct 23, 2026; Xinhua English summary, Oct 23, 2026.

Related coverage

Cite this article

NeuroAI (脑机网), "China Issued 18 Measures to Make AI Adoption Friendlier to Jobs," 19 November 2026, https://neuroai.site/a/na-policy-state-council-ai-employment-guide

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