China Orders Steps to Soften AI Job Displacement

China’s government wants artificial intelligence to create work faster than it removes it, and it has now written down how. A guideline published on 23 October tells local governments and ministries to build new AI industries and occupations, retrain workers, and slow the use of AI where it would hit jobs hardest.

The document is unusually direct about displacement. It names AI’s “job-substitution impact” as something to be mitigated (translated), asks companies to warn workers before large AI rollouts, and folds AI job risk into the government’s system for handling large-scale unemployment. It sets out 18 tasks across six areas but contains no figures, targets or deadlines.

In brief

What happened: The State Council General Office published guidance on adapting to AI to build an “employment-friendly development approach” (translated), with 18 tasks in six areas.
Why it matters: It pairs AI job creation and reskilling with job-risk checks, advance notice for workers and unemployment monitoring.
What’s next: Labour and planning departments are told to co-ordinate delivery; no timetable is given.

What the guideline sets out

The full text on the government’s website was issued by the State Council General Office with the State Council’s approval, as document Guobanfa [2026] No. 29. Xinhua, which published a summary at 10:02 BST on 23 October, said it had been issued “recently” (translated).

The first group of tasks is about creating work. Eastern regions are encouraged to host core AI industries, while central and western regions take on linked work such as data labelling and smart-device manufacturing. State funds, including the National Venture Capital Guidance Fund and the National AI Industry Investment Fund, are to back AI companies on market terms. The text also calls for faster recognition of new occupations, naming the AI trainer, the AI engineering technician and the digital management specialist.

Skills come next. Universities and vocational schools are asked to add AI majors and short “micro-majors” (translated), and local governments to steer graduates and other young people, migrant workers and unemployed people towards AI training. National online platforms are to offer more free public courses.

How it handles job losses

The displacement measures are the most specific part of the text. AI is to be deployed first in jobs that are short of workers or hazardous. In labour-intensive industries, officials should “prudently manage the scope, intensity and pace” of AI use, “minimising the impact on workers as far as possible” (translated).

Companies are encouraged to explore an “unemployment buffer period” system (translated), telling affected workers in advance when AI is rolled out at scale and offering help to move into new roles. AI projects in market-access areas that could cause large job shocks are to undergo a job-risk assessment when they are reviewed. Workers displaced by AI should get career guidance, training and job matching, and eligible insured unemployed people should receive unemployment insurance “in full and on time” (translated).

The guideline also covers rights at work. It calls for action against hiring discrimination caused by the abuse of algorithms, fake recruitment, and violations of privacy and rest rights, and for wider pilots of work-injury cover for platform workers. Finally, it orders job-risk monitoring, special surveys and new research centres, and puts AI job risk inside the mechanism for preventing and handling large-scale unemployment.

Reuters, via The Economic Times, summarised it this way: “China is urging government agencies to effectively mitigate job displacement caused by AI applications, according to guidelines issued on AI and employment on Friday.”

Most of the worker protections are encouragements, not obligations

The wording matters for anyone running AI projects in China. The unemployment buffer period, the clearest protection in the text, is something companies are “encouraged to explore” (translated). The same applies to much of the job-creation agenda: regions are encouraged, funds are steered and schools are asked. The guideline sets direction; it does not create a legal duty with penalties attached.

The job-risk assessment is the exception that deserves attention. For AI projects in market-access areas that could cause large job shocks, the assessment is to happen alongside the review itself, and the relevant departments are told to plan support for affected workers. That ties a labour test to an approval process, which is a firmer lever than encouragement. The text leaves open which projects count, who carries out the assessment and what happens if the result is poor.

For employers, the practical reading is that large-scale automation in China now carries a stated expectation of notice and transition support, even where no rule yet requires it. Firms planning big AI deployments in labour-intensive operations may want their workforce plans ready before regulators ask, since the guideline tells human resources and social security departments to co-ordinate delivery. Other governments are taking firmer routes: South Korean lawmakers have proposed making firms report AI-driven mass layoffs, and Italy has banned fully automated decisions on workers.

Beijing is treating AI displacement as an unemployment risk before publishing any numbers

The guideline does not deny that AI will cost jobs. It names the “job-substitution impact” (translated), asks for prudence in labour-intensive industries and places AI job risk in the same system the government uses for large-scale unemployment. That is a plain admission that losses are expected somewhere, even as the document leads with job creation.

What it does not offer is a measure of scale. There are no estimates of jobs at risk, no funding figures for retraining and no targets for new occupations. The monitoring tasks, including surveys of how AI is spreading, which jobs it creates and replaces, and how skill needs are changing, suggest the evidence base is still being built. Until that research produces numbers, the balance between the job-creation half of the plan and the displacement half cannot be judged from outside.

The reskilling side has the same gap. Training for graduates, migrant workers and unemployed people, free public courses and new university programmes are all named, but the text does not say how many people they should reach or by when. That makes the plan closer to New York’s AI jobs institute in ambition than in detail, since New York at least gave a training target. The test of China’s approach will be whether its local governments turn these headings into funded programmes, and whether the promised monitoring is ever published.

What happens next

The guideline tells local governments and departments to implement it in light of their own conditions, with human resources and social security, and development and reform departments leading co-ordination. No timetable is given. The first concrete signals to watch for are the new occupation standards it calls for and the first job-risk assessments on AI projects, the one measure with an approval process attached.