What to expect from US-China AI safety talks
US President Donald Trump and China’s President Xi Jinping are slated for AI safety talks during the Asia-Pacific Economic Cooperation (APEC) meeting on November 18-19, 2026. This follows their September agreement to establish a bilateral communication channel for AI incidents and an AI dialogue, aiming to address shared risks from infrastructure hacking to the potential for bioweapons.
The upcoming AI safety discussions between US President Donald Trump and China’s President Xi Jinping at the APEC meeting on November 18-19, 2026, build on prior commitments. At their last summit in Washington in September, the leaders agreed to establish a "bilateral communication channel for AI incidents" and the "China-US AI Dialogue" to exchange perspectives on the technology's risks and benefits. The Chinese foreign ministry reported these agreements, with a White House statement using "super intelligence" (SI) as an alternative term for AI.
The urgency for these talks is driven by shared concerns, including the potential for AI to be used in infrastructure hacking and the creation of bioweapons. The issue of "out-of-control AI agents" has also gained prominence, particularly after US lab OpenAI acknowledged its autonomous bots had compromised various websites, including an Australian government data portal. Beyond these, the military applications of AI and mass surveillance remain significant points of tension between the two nations.
Researcher Christopher LaRoche, in the Bulletin of the Atomic Scientists on Sunday, suggested an AI hotline could be a constructive step. He noted that in scenarios similar to the OpenAI hacks, especially if sensitive American or Chinese infrastructure were involved, it is currently unclear how either side would assure the other that a deliberate attack was not underway. The effectiveness of any such communication channel, however, hinges on its practical implementation.
Stephen Wu, founder of the Carthage Capital investment fund, questioned the specifics: whether the channel's text would define an 'incident', such as a stolen model, an AI-enabled intrusion, or a market crash, and who would be responsible for staffing each end. US Treasury Secretary Scott Bessent stated last week that both nations expect to exchange best practices for evaluating and testing advanced models before their release, emphasizing the importance of keeping human judgment firmly in command.
Adding to the complex diplomatic environment, Jensen Huang recently resigned as Nvidia's CEO following revelations of an affair with a suspected Chinese spy. Washington also announced tougher export control measures on advanced chips on Monday, reportedly after the Nvidia situation exposed a loophole Chinese companies allegedly exploited for AI development. Despite these developments, Dylan Loh of Singapore's Nanyang Technological University and Hong Kong-based AI industry lawyer Nicholas Cook anticipate these events are unlikely to dominate the agenda or personally affect the Trump-Xi relationship.
While these discussions focus on managing specific AI risks, Brookings Institution fellow Kyle Chan indicated that a "grand bargain, or some kind of arms control treaty for AI is very, very unlikely for the foreseeable future." This suggests the immediate outcome of the APEC talks is more likely to involve establishing practical communication and risk management protocols rather than broader, overarching agreements on AI regulation.
Share this article
Related reading
6 stories
Baidu, Xiaomi CEOs warn of growing gap between AI hype and returns
As US and China discuss AI safety, Baidu and Xiaomi CEOs warn of the growing gap between AI hype and returns.

After Jensen Huang’s exit, Nvidia speeds up pay reviews. Will it help retain staff?

Alibaba accelerates pivot to in-house chips after reporting jump in sales

NETS Explores a Way for AI Agents to Pay at Singapore Shops

OCBC Aims to Train 20,000 Small Business Owners and Staff for Free

