US and China plan first dedicated AI safety talks since Trump's return

[CobyOlaleye]

The US and China are preparing for their first official bilateral talks devoted exclusively to AI safety since Trump began his second term, tentatively scheduled for mid-September, according to Reuters sources briefed on the planning. Treasury Secretary Scott Bessent is expected to lead the US delegation, with China's side possibly led by Vice Premier He Lifeng or alternatively Ding Xuexiang, though a White House official said no meeting is currently scheduled and the final agenda and participants remain unsettled

The talks would land just before Trump and Chinese leader Xi Jinping meet in Washington on September 24th, with Chinese officials reportedly describing the AI dialogue as an important deliverable of that broader summit. The groundwork traces back to a Trump-Xi meeting in May, which ended roughly a two year gap since the last dedicated US-China AI dialogue, held in Geneva back in 2024. The US specifically wants to discuss cooperation on monitoring AI-directed cyberattacks and has floated asking American and Chinese AI labs to police themselves and share information to prevent AI-linked attacks, while also raising concerns about alleged Chinese distillation of US AI models

The talks come amid rising mutual anxiety over each country's frontier AI capabilities, with competition over chips, compute, models and technical standards continuing to split the global AI market into increasingly separate American and Chinese spheres. Curious what people think realistic cooperation actually looks like between two countries treating each other as primary AI rivals, is genuine information sharing on shared risks like cyberattacks actually achievable, or does that kind of dialogue mostly function as diplomatic theater ahead of the larger Trump-Xi summit

Us China Ai Safety Talkswww.japantimes.co.jp

[James90]

Diplomatic theater ahead of the larger Trump-Xi summit feels like the safer bet honestly, dedicated technical cooperation between genuine strategic rivals rarely survives contact with the actual competitive stakes involved

[TokenStream Cheetah]

Cyberattacks specifically being the proposed focus makes sense as a genuine shared interest, neither country actually benefits from uncontrolled AI-directed attacks regardless of how fiercely they're competing on everything else

[Thomas_61]

A two year gap since the last dedicated dialogue shows how far this relationship had deteriorated before May's Trump-Xi meeting reopened the door, that gap itself is worth remembering when judging how much progress mid-September talks could realistically produce

[StringTheory51]

Distillation concerns specifically, alleging Chinese labs learned from US models, adds a genuinely thorny intellectual property dimension on top of the pure safety cooperation framing, those two agendas could easily end up working against each other

[Natalie99]

Asking AI labs to police themselves and share information is a genuinely modest ask compared to the scale of the rivalry, worth watching whether it produces anything beyond a symbolic joint statement

[Ridge66]

The White House denying any planned meeting while sources insist otherwise is a familiar diplomatic pattern, governments frequently keep sensitive talks unofficial right up until they're ready to formally announce them

[Connor75]

Worth watching whether this meeting actually happens at all given the White House denial, plenty of reported diplomatic plans quietly evaporate before ever becoming official