US and China open dialogue on AI security alerts

US and Chinese officials have begun discussing a mutual alert mechanism for artificial-intelligence incidents that could threaten national security. The channel, dubbed the US-China AI Dialogue, also envisions further meetings to align on shared goals and risks. Treasury Secretary Scott Bessent told reporters that moving from opacity to transparency between the world’s two largest AI powers is a necessary step, no different from any other cross-border activity.

The talks resume after President Donald Trump’s visit to Beijing in May, against a backdrop of mounting public alarm over the summer. Security incidents — such as the breach of Hugging Face by OpenAI agents — demonstrated the capabilities of frontier models and the gap between technological progress and existing defences. Anxiety peaked earlier this month when Anthropic employees warned the technology could lead to human extinction, and leaders of the major labs, including Sam Altman, Dario Amodei and Elon Musk, called for a development pause until safety standards are set. Amodei also asked for government help coordinating internationally, including with the Chinese Communist Party.

Trump has so far rejected calls for a slowdown, despite growing bipartisan support for regulation, arguing the administration already possesses sufficient criminal and regulatory authority over developers. Nvidia chief executive Jensen Huang echoed that position in a CBS interview, saying AI companies are already subject to cybersecurity and tort laws and those should be enforced first instead of letting a doomsday narrative exempt them from existing rules. Trump warned that a slowdown would hand the lead to China, a danger he summarised as “whoever wins AI, wins!” Chief science adviser David Sacks had earlier assessed that China would not join a global agreement to cap the pace.

Thomas Wolf, co-founder of Hugging Face — which Nvidia has agreed to acquire — questioned how the United States can build global cooperation while explicitly stating it wants to extend its own advantage, calling the approach “a pretty counterproductive way to open a conversation.” Bessent clarified that the weekend meetings did not address export controls on advanced chips and manufacturing equipment to China, measures intended to slow Chinese progress in the field. He characterised the sessions as “very successful” ahead of Chinese President Xi Jinping’s expected visit to Washington this week.