US and China Begin Talks on Warning Each Other of AI Security Threats
US and Chinese officials have started discussions on a mechanism to notify each other about AI incidents that could threaten national security. The move comes amid growing global concern over AI safety failures and calls to slow the technology's development.
American and Chinese officials have opened talks on a proposed framework, called the US China AI Dialogue, that would let both countries alert one another to artificial intelligence incidents posing risks to national security. US Treasury Secretary Scott Bessent said the arrangement would also involve further meetings between representatives of both nations to discuss the technology and work toward shared goals and threat assessments.
Bessent said moving from opacity to greater transparency between the world's two leading AI powers is important, given how cross-border activity typically works. He noted the talks resume discussions that began after President Donald Trump's visit to Beijing in May, a period during which public concern over AI safety has intensified.
Several incidents over the summer fueled that concern, including a case where OpenAI's AI agents hacked into Hugging Face, exposing gaps between the capabilities of leading AI models and the industry's safety safeguards. Alarm grew further this month after employees at Anthropic warned that AI development could be heading toward catastrophic, human-threatening outcomes.
In recent weeks, leaders of major AI labs — including OpenAI's Sam Altman, Anthropic's Dario Amodei, and Elon Musk — have called for slowing AI development to allow time for safety standards to catch up. Amodei, in an essay outlining possible steps for labs and governments, called for government help in building international coordination, including engagement with the Chinese Communist Party.
Trump has largely dismissed these appeals so far, even as bipartisan support grows for regulating the industry; his administration argues existing criminal and regulatory powers are already sufficient. Nvidia CEO Jensen Huang echoed a similar view, saying AI firms are already bound by cybersecurity and liability laws that should be enforced before new rules are added.
Trump has warned that slowing down AI development risks ceding US leadership to China. His science adviser David Sacks has said China is unlikely to join any global agreement to pace AI development. Hugging Face cofounder Thomas Wolf questioned whether the US can credibly pursue global cooperation while openly seeking to extend its own technological lead.
The US continues to restrict exports of advanced chips and chipmaking technology to China as part of efforts to slow its AI progress, though Bessent said these export controls were not part of this weekend's discussions, which also covered trade matters ahead of Chinese President Xi Jinping's visit to Washington this week.
