China issues AI ethics and safety guideline 1.0 for big models

This guideline is Beijing’s strongest signal yet that the era of free‑wheeling large‑model deployment inside China is over. By turning high‑level “AI ethics” rhetoric into a concrete lifecycle standard—covering training data, deployment, hallucination control and user‑side safeguards—the state is telling labs and platforms what “responsible” looks like in operational terms. The involvement of Alibaba, Huawei and DeepSeek in drafting the document indicates that top labs expect to live under this regime and have shaped it to be implementable rather than purely aspirational.([aibase.com](https://www.aibase.com/zh/news/28313))

For the AGI race, this moves China closer to a European‑style, standards‑driven governance model rather than the US’s litigation‑driven approach. In the short run, compliance overhead may slow some product launches and add friction for smaller players. Over the medium term, however, it could actually derisk aggressive deployment of very powerful models inside critical infrastructure, healthcare and finance. If Chinese labs can show regulators that systems meet a state-blessed safety bar, it may accelerate roll‑out of capabilities that would be politically sensitive elsewhere.