China will screen AI projects for job losses under a new State Council plan

What China published

China's State Council has told local governments and ministries to weigh job losses before backing AI projects and to build early-warning systems for AI-driven unemployment. The General Office of the State Council published the full text of the opinion on the central government website on 23 October 2026. Its document number is Guobanfa 2026 No. 29, and it is dated 16 October 2026. Xinhua summarized it the same day, in a report carried by China News Service.

Its Chinese title translates roughly as Opinions on Adapting to the Development of Artificial Intelligence Technology and Helping Build an Employment-Friendly Mode of Development. It sets out 18 tasks under six headings, plus a general requirements section and an implementation section. It is addressed to provincial governments and to State Council ministries and agencies, and was issued with the council's approval.

The goal: an employment-friendly way to grow

The document opens with a standard line in Chinese policy writing: employment is the most basic part of people's livelihoods. Its stated aim is to strengthen the positive effects of AI on creating jobs, upgrading traditional roles, improving job quality and encouraging entrepreneurship, while effectively easing what it calls the job substitution impact of AI applications.

The framing is balance rather than restriction. The text wants technological progress and industrial upgrading to move together with job creation and better living standards, which it calls inclusive development. It also repeats the government's employment-first policy orientation and tells every region and department to take the employment effects of AI into account in their own work.

Job risk reviews for AI projects

The most concrete new tool is task seven, an employment risk assessment mechanism for AI applications. Where an AI application scenario project in a market-access field could cause a large employment shock, the opinion says a job-risk assessment should be carried out at the same time as the project is reviewed. The relevant departments are then expected to arrange support for affected workers to change jobs or roles. For AI projects in other fields, the assessment should be done where circumstances warrant.

The same task also asks officials to assess AI's overall effect on jobs and to steer innovation resources toward areas with the most potential to create work. The document does not define what counts as a large shock, which projects fall into the market-access category, or who carries out the assessment.

Slowing automation in labour-heavy industries

Task six goes further in tone. It says AI should first be deployed where labour is scarce or the work is dangerous, where it can improve working conditions and safety. In labour-intensive industries, governments and firms are told to handle the scope, intensity and pace of AI adoption steadily and cautiously, to reduce the impact on workers as much as possible.

Companies are encouraged to explore an unemployment buffer period system. When they roll out AI at scale, they should tell affected workers in advance and offer support to move into new jobs or roles. The text encourages this rather than requiring it, and gives no length for the buffer period.

Task five asks for human-machine collaboration in the AI upgrade of industry and farming, and for service industries to absorb more workers as AI reshapes them.

Creating AI jobs and new occupations

The first group of tasks is about growth. The opinion tells the country to keep implementing its AI Plus action plan, encourages eastern regions to build core AI industries, and guides central and western regions toward related work such as data labelling and smart device manufacturing. Eligible AI firms and specialized small companies can be included in special bank loan programmes meant to stabilize and expand employment.

It also calls for an AI entrepreneurship support plan for small companies, low-cost lightweight digital tools, and equal legal access to data and algorithms for small businesses and sole traders. Government investment funds, named as the national venture capital guidance fund and the national AI industry investment fund, are to back AI firms on market terms.

The document wants a high-level AI open-source community and a batch of quality open-source projects, as a way to create startups and jobs. That sits alongside China's other recent AI policy moves, including Commerce Ministry plans for export controls on 23 AI technologies. Task four asks for faster creation of new occupations and job standards, naming digital management specialists, AI engineering technicians and AI trainers, and for new professional title tracks.

Training, from universities to migrant workers

Universities and vocational schools, including technical colleges, are told to adjust their programmes quickly, add urgently needed AI disciplines on an extraordinary scale, and offer more AI micro-majors. Companies should help with course design and teacher training, engineers should teach in schools, and students should do placements in companies. AI education is to cover all school stages.

A special AI training plan targets graduates and other young people, migrant workers and the unemployed. Public training bases and company training centres are to run AI skills courses, national online learning platforms should offer more free classes, and existing vocational skills competitions should add AI events.

Rules for AI in hiring and pay

Task thirteen covers workers' rights. It calls for research into basic protections suited to human-machine collaboration and for companies to be supervised in paying wages, respecting rest and holidays and providing workplace safety. It says the use of AI in human resources management should be regulated, and that illegal conduct such as employment discrimination caused by abusing algorithms, fake recruitment, and violations of privacy and the right to rest should be dealt with according to law.

That language comes as other governments write their own rules for AI in the workplace. Italy, for example, now voids dismissals decided by AI alone. China's text points to existing law rather than creating a new ban.

On social insurance, the opinion encourages gig and flexible workers to join the basic pension scheme for urban employees and wants to expand pilot occupational injury cover for platform workers. On pay, it asks companies to build salary systems suited to AI staff, apply minimum wage guidance for skilled workers, and invest more in public-interest AI, services and open-source communities. It also names intelligent agents as an emerging form of work to be developed.

Monitoring and a layoff playbook

The final group of tasks builds a monitoring system. Ministries are to share data on AI industries and employment, and to build regional and sector-level models that analyse job data. A special survey will track how far AI has spread, how many jobs it creates and replaces, and how skill demand changes. A set of AI employment research centres will be set up at universities and research institutes.

Task eighteen brings AI-driven job losses into China's existing mechanism for preventing and defusing large-scale unemployment, and calls for contingency plans and joint handling of labour disputes linked to AI. Public employment services, including the national job information database and graduate job platform, are meant to use AI themselves to match people with support. Laid-off workers who are insured should receive unemployment benefits in full and on time.

Human resources and social security departments and development and reform departments at every level are told to coordinate the work. Major issues go to the Party Central Committee and State Council.

What the document does not say

The opinion contains no numbers. It sets no target for jobs created, no threshold for the risk reviews, no budget and no deadline. It does not name sectors where AI should be slowed, beyond the general reference to labour-intensive industries, and it creates no penalty for companies that automate quickly. Much depends on detailed rules that ministries and provinces have not yet published.

It also gives no figures on how many Chinese workers have already been affected by AI, and the Xinhua summary adds none. Readers should treat it as a statement of direction from the cabinet, with the practical effect to be judged when implementing rules appear.

How it compares

Few governments have gone as far as tying job-risk reviews to the approval of AI projects. In the United States, the policy argument has focused more on investment and competition, such as the Treasury proposal to ban US investment in superintelligence work in China, than on labour planning. Company data is starting to fill the gap; an Anthropic study found that agent hours outnumber staff hours six to one in some settings, as junior hiring fell.

China's document treats job losses from AI as a risk to be managed through the state's existing employment and unemployment machinery, while still pushing hard on AI adoption. Whether the job-risk reviews end up slowing any real project is the part to watch.