China's State Council General Office has issued a policy opinion on AI and employment that calls for an AI employment risk assessment mechanism and encourages advance notice to affected workers

Key takeaways

• The opinion is a policy document (in our reading) with 18 measures, addressed to governments and State Council departments.
• Measure 6 encourages firms to explore an unemployment buffer period and to give affected workers advance notice.
• Measure 7 calls for an AI employment risk assessment mechanism, run alongside review for market-access AI scenarios likely to cause a large job impact.
• Measure 13 says AI in HR management should be regulated and algorithm-driven discrimination and fake recruitment dealt with according to law.

Is the State Council opinion on AI and employment binding?

In our reading it is a policy opinion, not a statute; the source does not label itself binding or non-binding. The document is Guo Ban Fa [2026] No. 29, dated 16 October 2026 and marked for public release, and it is addressed to the people's governments of provinces, autonomous regions and municipalities and to State Council ministries and agencies. Its closing part tells regions and departments to put it into practice, with human resources and social security and development and reform departments coordinating. It states no penalty and no effective date.

Most employer-facing items are worded as encouragement or as application of existing law. A comparable agency-level instrument on a related technology is summarised in our article on the CAC, NDRC and MIIT opinions on AI agents.

What does it say about advance notice and a buffer period for workers?

Measure 6 is the passage most relevant to employers. It says AI should first be promoted in roles with labour shortages or high-risk environments, and that the scope, intensity and pace of AI use in labour-intensive industries should be handled steadily and prudently to limit the impact on workers. It then states: "鼓励企业探索建立失业缓冲期制度" (the State encourages firms to explore an unemployment buffer period), so that affected workers are told in advance during large-scale AI rollouts and given transition and re-placement support.

The wording is "encourage". The source does not define the length of a buffer period, does not say who enforces it and sets no consequence for skipping it. Measure 12 separately says that, as the rules provide, eligible workers affected by AI job substitution should receive employment assistance, and that relevant insured unemployed persons should receive unemployment insurance benefits promptly and in full.

How would the AI employment risk assessment work?

Measure 7 is titled "建立人工智能应用就业风险评估机制". It says the employment effect of AI applications should be assessed overall, with innovation resources steered toward uses that create jobs. For AI scenario projects in market-access fields that could cause a large job impact, the employment risk assessment would be carried out during the review itself ("在评审时同步开展就业风险评估"), and relevant departments are to coordinate transition support. For AI scenario projects in other fields, the assessment is to be done in light of actual circumstances.

The source does not list which scenarios count as market-access fields, does not set a threshold for a large impact and does not name the reviewing bodies. Measures 16 to 18 add monitoring, a special survey on AI and employment, and a response mechanism that would bring AI employment risk into mass-unemployment prevention plans ("将人工智能应用就业风险纳入规模性失业风险防范化解机制应对范围").

What does it say about AI in HR management and hiring?

Measure 13 says AI use in human resources management should be regulated, and that unlawful conduct from algorithm misuse should be dealt with according to law, naming "依法治理因滥用算法等造成的就业歧视、虚假招聘" (employment discrimination and fake recruitment) together with violations of privacy and rest rights. It also asks departments to see that firms secure workers' pay, rest and leave, and safety and health protections.

The measure points to existing law and does not create a new offence or cite a specific statute. Measure 11 takes a different tone toward the sector: it encourages operators to develop recruitment large models, AI interviews and virtual-reality training products. For comparison with US state rules on automated hiring tools, see our piece on the four-state AI hiring rulebooks, which applies in a different system.

What should employers and HR teams in China do now?

The opinion does not require these steps. They are the desk's suggestions for groundwork that holds up whatever implementing rules follow, if any.

• List the roles and teams in China where AI is changing headcount, duties or working hours, and note the planned pace.
• Write a short job-impact assessment for any large-scale rollout, and keep it with the project approval record.
• Draft a notice and transition-support plan for affected workers, and have local counsel check it against labour contract law.
• Review AI screening, interview and scheduling tools for discrimination, privacy and rest-time risks, since measure 13 names them.
• Watch for implementing rules from the human resources and social security and development and reform departments.

What we did not verify

What we opened: the full Chinese text of the opinion on gov.cn, including its header block (document number, dates, issuing body), on 24 October 2026.

What we did not open: any implementing rule, any official explainer (the page shows an interpretation heading with no item listed, so there was none to read), the local rules of any province, and the text of the labour laws the opinion refers to. English renderings of the Chinese text are ours.

What we refuse to claim: that employers are required to give notice or adopt a buffer period, that any penalty exists, that specific ministries have issued follow-up rules, or any effective date.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• HR and employment: AI regulation
• China opinions on AI agents from CAC, NDRC and MIIT
• China Supreme People's Court opinions on adjudicating AI disputes
• AI hiring: four US state rulebooks, one workflow
• California CPPA automated decision rules and employment decisions