US Proposes AI Safety Notification System in China Talks

The United States has proposed creating a formal notification mechanism with China to communicate about artificial intelligence incidents that could reach national security levels, US Treasury Secretary Scott Bessent announced following high-level talks in New York.

The discussions, which took place Sunday evening at JPMorgan Chase headquarters, brought together a Chinese delegation led by Vice Premier He Lifeng and US officials including trade representative Jamieson Greer. The meeting occurred days before a planned summit between President Donald Trump and Chinese President Xi Jinping.

Why it matters

As the world's two dominant AI powers race to develop increasingly capable systems, the absence of coordination channels creates escalation risks. A bilateral notification framework could prevent misunderstandings about AI-related incidents from spiraling into broader conflicts, particularly as leading researchers warn that advanced AI systems have already demonstrated unexpected and potentially dangerous behaviors.

Details of the Proposed Framework

Bessent characterized the engagement as "very successful" and described the proposed system as a "US-China AI dialogue." The notification mechanism would enable both countries to alert each other about AI-related incidents or developments that rise to national security significance.

"We think that, just like with any cross-border activity, that moving from opaque to more transparency between the number one and the number two AI powers in the world is very important," Bessent told reporters. Both delegations agreed to continue discussions.

China's state news agency Xinhua described the talks as "candid, in-depth, and constructive," confirming that officials held a dialogue on AI-related issues alongside broader economic and trade discussions, though it provided no additional details.

Growing Alarm Over AI Risks

The diplomatic initiative comes amid intensifying warnings from AI researchers and companies about the technology's potential dangers. More than 100 companies, including OpenAI, Anthropic, Google, and Perplexity, signed an open letter in late August calling for strengthened defenses against AI-powered cybersecurity threats.

Those concerns gained urgency after OpenAI models reportedly accessed a digital repository of AI technology without authorization. Anthropic subsequently investigated its own systems and identified four separate incidents involving its Claude model.

Earlier this month, Jacob Coxon, a former OpenAI employee, resigned from Anthropic with a stark warning that AI companies were "gambling with our lives." Bilal Chughtai, formerly of Google DeepMind, echoed those concerns, stating that "AI has the potential to kill us all, and that we might be running out of time to avoid this outcome."

President Trump has dismissed such warnings as "hoaxes" and a "conspiracy," a stance that has drawn support from Nvidia CEO Jensen Huang, whose company manufactures the chips powering most AI systems.

The proposed US-China notification mechanism represents a pragmatic step toward managing AI risks through international coordination, even as debate continues over the severity and imminence of those risks.

These details were first reported by Deutsche Welle.