Safety-risk disclosure and incident reporting could gain a statutory basis for companies building or running AI in China, with sandboxes and transition periods softening the shift.
China’s National Development and Reform Commission (NDRC) will speed up work on a basic, comprehensive AI law, Wang Ruomeng, a deputy director-general in the planner’s innovation and high-tech development department, told a press conference on November 3, according to the NDRC’s transcript. State news agency Xinhua reported that the plan also includes safety standards and ethical guidelines covering the full life cycle of AI and regulatory sandboxes for different sectors.
The pledge itself is not new. The State Council’s 2026 legislative work plan, published on May 11, already called for faster work on comprehensive legislation for the healthy development of AI, and the National People’s Congress Standing Committee’s plan for the year asked for research and drafting, with deliberation to be scheduled as appropriate, Legal Daily reported. In July, NDRC spokesman Jiang Yi said the planner would speed up the legislative process for an AI law, IT Home reported.
What is new is the detail. The briefing was part of a series explaining how the 15th Five-Year Plan, which runs from 2026 to 2030, will be carried out, and Wang set out the building blocks the planner wants in the law. He did not provide a draft, a timetable or the name of the body that will write it, but the list gives companies a preview of the obligations likely to be debated once a draft appears.
The planner listed disclosure, sandboxes and transition periods as the core tools
According to Beijing Daily, in a report carried by Sina, Wang described the aim as building the main pillars of China’s AI legal system. He said the government would set reasonable transition periods for new regulatory policies and make supervisory requirements and enforcement standards transparent and predictable.
The most concrete obligation in the remarks concerns companies. Wang said developers and operators should disclose safety risks in time and proactively report safety incidents and vulnerability information. He framed safety as a shared responsibility of government departments, industry groups, companies and the public.
Sandboxes are meant to cushion that burden. Wang said China would support AI sandboxes aimed at different fields, with mechanisms for application, testing, use and exit, preset plans to absorb risks, and more room for trial and error, Beijing Daily reported. Shanghai Securities News reported that the planned laws would cover data, computing power, algorithms and key application scenarios, with periodic reviews to adjust the rules.
Job displacement and harmful uses are also inside the planned framework
Employment featured prominently. Wang said the government would strengthen monitoring and early warning of the risk that AI replaces workers, link the social security system more closely to changes in employment structures and ways of working, and provide basic, universal social protection, according to a China.com report credited to Jiemian. Xinhua reported that new occupations such as model trainers and data annotators would be pushed into the national occupational classification. EastFrontier has followed earlier steps in this area, including an AI employment action plan whose standards and service tools were still being prepared.
China National Radio reported that Wang also promised stronger traceability against algorithmic discrimination, deepfakes and privacy leaks, special protection for older people and minors, and measures against excessive use of and emotional dependence on AI. He called AI safety a global challenge that no single company, industry or country can solve alone.
The legal agenda sits beside an expansion program. Wang said the "AI Plus" initiative would prioritize dangerous and difficult work such as mining, work at height and emergency rescue, and that national AI pilot bases in manufacturing, energy, transport and healthcare would continue, CGTN reported. At the same briefing, Ren Zhongbao, another deputy director-general in the department, said China’s emerging pillar industries produced nearly RMB6 trillion last year and could exceed RMB10 trillion by the end of the plan period, the 21st Century Business Herald reported.
A draft and its lead drafter are the next things to look for
For companies, the useful reading is the list of obligations. Risk disclosure and incident reporting would give AI providers a formal reporting line to regulators backed by statute, while sandboxes and transition periods suggest the government wants to avoid freezing deployment while rules are written.
Several questions remain open. The briefing did not say which body would lead the drafting, how the law would treat open-weight models, or how the disclosure duties would apply to companies serving users abroad.
With the law already on both legislative plans, the clearest signal will be a draft: one submitted to the NPC Standing Committee for deliberation, or released for public comment. Until then, the sector-specific sandbox pilots that Wang described are the place where the planned rules will first be tested on real services.