Chinese labs are wild cards in AI governance
BABA•Chinese AI labs Z.AI and Alibaba are less open about safety than OpenAI and Anthropic, partly because of chip and capital shortages, and will need to improve to retain access to markets from Brussels to Seoul. Z.AI and Concordia AI released what they called the first comprehensive AI safety framework from a Chinese company on September 28.
1. Safety disclosures differ
OpenAI and Anthropic publish risk reviews ahead of model releases, while Z.AI and Alibaba are less open on safety, partly because of a shortage of chips and capital. The column says Chinese labs will have to improve their practices to retain access to markets from Brussels to Seoul.
2. Z.AI safety steps
Z.AI and Beijing-based consultancy Concordia AI released an open-weight model risk management report on September 28, described as the first comprehensive AI safety framework from a Chinese company. On September 21, Z.AI said it had disabled some features of its flagship coding assistant after users reported it was uploading entire local code repositories to overseas cloud servers without their consent. In August, Z.AI released GLM-5.3 after a two-week review, saying it neared Anthropic's Mythos in finding software vulnerabilities; it was the first Chinese lab to explicitly delay a model release for safety reasons.
