Despite trust deficit, US and China explore alert system for AI safety
The United States and China will hold further talks on AI safety, despite ongoing competition. US Treasury Secretary Scott Bessent announced a new US-China AI dialogue on September 20. He proposed a notification system for AI incidents that could threaten national security. This follows accusations by three US government agencies on September 8 that six Chinese firms, including DeepSeek and Moonshot AI, engaged in "malicious campaigns" of AI output distillation.
The US and China are talking about AI safety, but the US push to formalize a notification system for AI incidents stands out. This comes as the US accuses Chinese firms of stealing American AI output. The dialogue itself is a minor win for stability, but the underlying tensions remain high. A formal incident alert system would be a significant step, even without full trust.
For Asia, this US-China dialogue is a test of whether competition can coexist with risk management. Moonshot AI and Alibaba, both named in US accusations, face scrutiny. China's stated commitment to AI safety is now directly challenged by US claims of "malicious campaigns." This creates a difficult path for any true cooperation, especially on sensitive issues like AI weaponization.
The thing to watch is the next meeting, potentially in two months. If the US and China can agree on a framework for the proposed notification system, it would show a genuine commitment to de-escalation. Without such a concrete step, the dialogue risks becoming a symbolic gesture, with accusations of distillation continuing to undermine trust.
Share this article
Related reading
6 storiesUS, China to meet again on AI safety in two months in Shenzhen, Bessent says
We previously reported on the US and China's ongoing AI safety discussions, including future meeting plans.

The most dangerous AI claims aren’t wrong — they’re overstated

I Built an AI Safety Layer on a 4GB Phone

Don't let the AI change its own pass condition: pick the protected files, stop on any change

Your SKILL.md is portable. Is its behavior? Introducing Skillporter

