China joins voices warning of AI self-improvement

China's Cybersecurity Standardisation Technical Committee (CSTC) published its annual AI Safety Governance Framework on 14 September, warning that AI models are showing a "self-accelerating trend of model and algorithm autonomous learning, optimization, and recursive self-improvement" that could "exceed human anticipation." The framework's call for "attention and vigilance" echoes concerns raised by Anthropic CEO Dario Amodei about rapid AI development, even as Chinese Foreign Ministry Spokesperson Guo Jiakun dismissed Amodei's call for an AI slowdown as "fear-mongering, confrontation and vicious competition.

While Chinese government officials criticised calls for an AI slowdown led by Anthropic CEO Dario Amodei, China’s cybersecurity administration said models were increasingly demonstrating “recursive self-improvement”. A new AI safety framework https://www.cac.gov.cn/2026-09/14/c 1791137092283345.htm?ref=thestack.technology published by the country’s Cybersecurity Standardisation Technical Committee CSTC on 14 September called for “attention and vigilance” on the issue that AI could “exceed human anticipation.” The annual AI Safety Governance Framework said: “AI has demonstrated a self-accelerating trend of model and algorithm autonomous learning, optimization, and recursive self-improvement.” The warning is similar to Amodei’s concerns about rapid AI development https://darioamodei.com/post/we-must-pace-the-frontier?ref=thestack.technology and contrasts with Chinese Foreign Ministry Spokesperson Guo Jiakun's statement that https://za.china-embassy.gov.cn/eng/fyrth/202609/t20260914 12021997.htm?ref=thestack.technology Amodei’s call for an AI slowdown was “fear-mongering, confrontation and vicious competition” that would only slow work on global AI governance.