AI Safety in China: A Primer - by Jasmine Li - The J-Space

A map of the landscape, & ways to contribute

2026 has heralded the summer of powerful Chinese models: Kimi K3, Qwen3.8, and Z.ai’s GLM-5.3 have arrived in quick succession, all open-weighted and exceedingly close to the closed-model frontier. At the same time, the AI risk frog pot is heating up globally. China is crucial in enabling humanity to navigate AI risks well, but the Chinese AI safety ecosystem is, in my experience, poorly understood. In this post, I: Provide an overview of China’s AI safety ecosystem, which I think is crucial context for thinking seriously about ‘win conditions’ for AI safety; Highlight concrete ways to…

saved by

• Alex Yun
• Arunim Agarwal

related reading

• How “AI Safety” Traveled Through the Chinese Internetyinwei.substack.com
• Who should the U.S. Talk to in China on AI?mattsheehan.substack.com
• China AI Bulletin 1 - by Emmie Hine - China AI Bulletinchinaaibulletin.substack.com
• Is China Serious About AI Safety? | AI Frontiersai-frontiers.org
• China Says It Cares About Loss of Controlocpl.substack.com
• China AI Bulletin 5substack.com
• Recommendations for Technical AI Safety Research Directionsalignment.anthropic.com
• Chinese Audiences Are Reading Western AI Safety Discourse | AI Frontiersai-frontiers.org
• Making Money in Chinese AI Safetychinatalk.media
• China’s AI Regulations and How They Get Madecarnegieendowment.org
• China’s AI Companies Are Going Closed Sourcechinatalk.media
• A Path Forward on AI Safety for the United States and Chinacarnegieendowment.org