What happened
On 3 November 2026, senior National Development and Reform Commission (NDRC) official Wang Ruomeng told a Beijing briefing that China will 'accelerate legislation' to support core AI enablers — data, computing power and algorithms — and regulate key application scenarios, while establishing safety standards and ethical guidelines covering the full lifecycle of AI. Developers and operators would be expected to strengthen risk management and disclose safety risks, incidents and vulnerabilities, with a transition period, guidance and sector-specific regulatory sandboxes to ease compliance, and continued encouragement of open AI development. No final rule text, effective date or penalty schedule was published.
Why it matters
This is a high-level statement of direction from China's economic-planning body confirming a shift from content/output-focused AI rules toward lifecycle standards covering training data, compute, evaluation and incident handling — closer in structure to the EU AI Act and US frontier-model proposals. It signals likely new mandatory safety/disclosure obligations for Chinese labs (e.g. DeepSeek, Moonshot, Alibaba) and, indirectly, for overseas teams relying on Chinese open-weight models (model cards, safety evaluations, incident reporting). No binding instrument yet.
Action needed
For organisations using Chinese open-weight or hosted models, monitor for subsequent CAC/NDRC/TC260 rulemaking and prepare to satisfy any future lifecycle safety and incident-disclosure requirements; track whether disclosure becomes a legal duty and who receives incident reports.