China will accelerate legislation covering artificial intelligence, including rules on data, computing power, algorithms and key applications, according to a senior official at the country’s main economic planning body.
Wang Ruomeng, a deputy director-general at the National Development and Reform Commission, set out the plan at a briefing in Beijing on 3 November.
What the plan covers
The legislation would establish safety standards and ethical guidelines across the full life cycle of AI models, from development to deployment and retirement. It would also introduce a transition period and sector-specific regulatory sandboxes, allowing companies to test systems under supervision before they are covered by the full rules.
Developers and operators would be required to disclose safety risks, incidents and vulnerabilities. That is a significant obligation, because it would make companies responsible for reporting failures that they might otherwise prefer to keep quiet.
The plan also commits the government to monitoring the effect of AI on employment and to strengthening safety-net protections for workers whose jobs are displaced.
Why China is legislating now
China already regulates some AI applications, including recommendation algorithms and generative services, through rules issued by several agencies. Those rules are fragmented, and companies have complained that they overlap and that the requirements are unclear.
A single law would consolidate the framework and give the central government a clearer instrument for directing development. It would also signal to foreign governments and investors that China intends to regulate AI systematically rather than case by case.
The timing follows a September agreement between China and the United States on AI and a summit between Xi Jinping and Donald Trump. Both countries have been developing governance frameworks while competing for advantage in model development and chip supply.
The balance the law must strike
China’s approach has been to regulate applications while encouraging development. The 15th five-year plan to 2030 treats AI as a strategic industry, and the new law is expected to include provisions that support open-source development and domestic chip production.
The risk is that disclosure requirements and safety audits slow the release of Chinese models at a time when American companies are shipping updates every few months. The transition period and sandboxes appear designed to manage that tension.
The employment provisions are unusual for a technology law. China’s leadership has become more explicit about the social consequences of automation, and the commitment to monitor job losses suggests that AI policy is now being written alongside labour policy rather than separately.
What to watch
The draft text will be published for consultation, and its scope will determine how much of the current patchwork survives. Companies will be watching the definitions of high-risk applications, the penalties for non-disclosure, and whether the law applies to open-source models distributed without a commercial service.
For foreign companies operating in China, the law will add a compliance layer that sits alongside data protection and cybersecurity rules. For Chinese developers selling abroad, it may become a marketing issue if the law requires disclosure of information that overseas customers would not otherwise see.
