Employment risk assessment upfront, unemployment buffer period written into document: State Council Office draws three lines for AI substitution

The General Office of the State Council recently issued the "Opinions on Adapting to the Development of Artificial Intelligence Technology to Help Build an Employment-Friendly Development Model", deploying 18 main tasks in 6 areas. This is the first time at the national level that the 'impact of AI applications on employment' has been systematically written into a special policy document: it not only discusses how to use AI to create jobs, but also how to ease the pain of substitution—risks must be assessed upfront, enterprises must provide a buffer period, and the pace of application in labor-intensive industries must be 'steadily and prudently grasped'. The focus of China's AI governance has shifted from simply 'promoting development' to 'promoting development + managing distribution'.

Policy Highlights: Three Mechanisms Appearing for the First Time

The document covers six major sections: job creation, upgrading paths, literacy improvement, employment services, rights protection, and monitoring and early warning. Among them, three mechanism designs are worth examining one by one:

• Employment risk assessment mechanism for AI applications. For AI application scenario projects in market access areas that may cause a relatively large employment impact, employment risk assessments are carried out simultaneously during review—risk prevention and control is moved forward from 'post-event remediation' to the 'access threshold'. This treats employment impact as an access threshold similar to environmental impact assessment.
• Unemployment buffer period system. Enterprises are encouraged to inform affected workers in advance and provide transition and job-transfer employment support during the large-scale promotion of AI. The original wording of the document is 'effectively mitigate the employment substitution impact of artificial intelligence applications'—not prohibiting substitution, but requiring that substitution must provide a process.
• The 'steady and prudent' principle for labor-intensive industries. Steadily and prudently grasp the scope, intensity, and pace of AI application in labor-intensive industries to minimize the impact on practitioners. The simultaneous appearance of three qualifiers is quite rare in industrial policy texts.

Substitution Rate Data: Why Policy Is Acting Now

The official interpretation cites Tsinghua University's "Skills Trend Report in the AI Era": legal document assistants 92%, junior programmers 87%, basic market analysts 85%—those under the greatest substitution pressure are precisely 'entry-level mental labor'. At the same time, the penetration rate of key AI scenarios in manufacturing above designated size has exceeded one-third, smart factories have been built in batches, and AI is changing from an option to a standard configuration. Both the supply side (model capabilities) and the demand side (enterprises' motivation to reduce costs) are accelerating, and what policy needs to close is the time gap: ensuring that workers' transition speed is not left behind by the speed of technological iteration.

The policy-oriented picture of human-machine collaboration: robots taking up posts in parallel with worker job-transfer training, and safety standards implemented simultaneously. (Illustrative image)

One Legal Precedent Drew the Bottom Line for Enterprises

Before the document was issued, judicial practice had already provided an answer: in a typical case in Hangzhou in May this year, a company's AI large-model quality inspection supervisor was dismissed on the grounds that 'AI can complete his work', and the court ultimately ruled that the company had illegally terminated the labor contract and ordered compensation of 260,000 yuan. The meaning of this line is very clear: enterprises cannot directly pass on the risks of normal technological updating to workers. The unemployment buffer period in the "Opinions" is essentially the extension of the principle established by the precedent into an institutional option.

But a buffer period is not a panacea. Advance notice is easy; where the job-transfer channels are, who will hire after training, and how to make up for the pay gap—these are the truly hard nuts to crack in 'mitigating substitution'. The supporting measures given in the "Opinions" are: implement special AI training programs, promote general education across society, accelerate the development of new AI occupations and release new occupational standards, and improve social security for flexible employment and workers in new forms of employment.

Same-Day Reinforcement: 16 National Standards for AI and Robots

The State Administration for Market Regulation approved and released 515 national standards in the same period, including 16 items on artificial intelligence, humanoid robots, and industrial environment robots, involving AI system safety and robot technical requirements. It is no coincidence that they were implemented on the same day as the employment policy: one governs 'what to do about people', and the other governs 'how to manage machines'. Together with 7 mandatory national standards including data security for intelligent connected vehicles, the governance framework for AI is moving from 'documents' to 'executable technical specifications'.

Three-Step Actions for Enterprise HR and Managers

• Inventory scenarios: sort out the business scenarios where AI is planned to be launched, conduct internal pre-assessments against the 'relatively large employment impact' standard, and retain assessment records—this is directly related to future access reviews.
• Design buffers: before large-scale deployment, determine the notification period, job-transfer training, and compensation plans, and write the 'unemployment buffer period' into the change management process rather than into a crisis public relations plan.
• Keep pace with standards: the 16 new national standards for humanoid robots and AI system safety will gradually become the basis for procurement and acceptance; comparing in advance can avoid secondary modifications.

Frequently Asked Questions (FAQ)