US and China discuss sharing alerts about AI security incidents

In short: US and Chinese officials have begun talks on a plan to notify each other about AI-related incidents that could threaten national security.

What happened

US Treasury Secretary Scott Bessent said the two countries are discussing a proposed framework called the US China AI Dialogue. One idea is to set up a process for each side to alert the other when an AI incident could pose a national security risk.

Think of it like an emergency phone line between two governments, but for major AI problems. The goal, Bessent said, is to move from “opaque” to more transparency between the two leading AI countries.

Bessent said the talks picked up after US President Donald Trump’s visit to Beijing in May. The discussions come after a summer of high-profile safety and security concerns tied to advanced AI systems, including incidents involving AI “agents” (tools that can take actions on a computer for a user, like a very fast assistant).

The idea of stronger safety rules has also been debated publicly by AI lab leaders, including OpenAI’s Sam Altman and Anthropic’s Dario Amodei, who have called for a slowdown while safety standards are set. Trump has largely resisted those calls, arguing that a slowdown could help China catch up.

Bessent said US export controls on advanced chips to China were not discussed in this weekend’s meetings. The meetings also included trade talks and took place ahead of a planned visit to Washington by Chinese President Xi Jinping.

Why it matters

If the US and China do build a working alert system, it could reduce the chance that a serious AI-related security incident spirals because one side learns about it too late. Even for regular people, these incidents can affect things like critical services, cyberattacks, and government decision-making.

Source: Wired