US and China could start warning each other about dangerous AI incidents under new proposal

The United States has proposed working with China to establish a system for alerting each other about potentially dangerous AI incidents that could threaten national security, according to senior US Treasury official Scott Bessent. The proposal emerged from weekend talks ahead of upcoming White House meetings between President Donald Trump and Chinese President Xi Jinping.

Bessent emphasized the importance of shared goals and awareness of shared threats, noting that increased transparency between the world's top two AI powers is crucial. However, the specifics of how the notification system would function remain unspecified, including which agencies would oversee it and which incidents would require an alert.

The agreement between the two nations to continue discussing AI safety measures is seen as a positive step, potentially setting an example for other countries to follow. This is one of several recent developments highlighting the need for better AI safety protocols. US and Chinese security experts have also advocated for a dedicated military hotline to prevent accidental or unauthorized AI actions from being misinterpreted as deliberate attacks.

Despite these calls, neither government has officially supported the idea. President Trump has consistently resisted calls for AI safety measures, dismissing them as a "hoax" and advocating for the US to lead in AI development both economically and strategically. Recent incidents involving AI systems, such as a false AI-generated intelligence report suggesting a Chinese vessel might be transporting nuclear materials, underscore the potential risks of AI misuse.

In one instance, the US nearly intercepted the Chinese vessel due to this error, with armed military personnel and aircraft en route to intercept the ship before the mistake was corrected.