The United States and China are preparing to sit down for their first dedicated AI safety dialogue since Donald Trump began his second term, with talks scheduled for mid-September 2026. The meeting, announced on September 4, signals a rare patch of cooperative ground between two countries that agree on almost nothing else right now.
US Treasury Secretary Scott Bessent will lead the American delegation. The talks are expected to focus on shared risks tied to AI technologies, with cyberattacks driven by autonomous AI systems near the top of the agenda.
How this came together
The groundwork for this meeting was laid at the Trump-Xi summit in May 2026, where both leaders agreed to restart formal AI discussions. That agreement ended a gap of roughly two years since the last dedicated US-China AI dialogue, which took place in Geneva in May 2024.
Chinese President Xi Jinping is scheduled to meet Trump in Washington on September 24, putting the AI safety dialogue just days before a broader bilateral summit. Beijing tends to treat technical working-level talks as scaffolding for higher-stakes diplomatic moments, and the proximity to the Washington summit suggests both sides see these AI discussions as something worth delivering results on before the leaders meet.
One notable design choice for the talks: AI safety is being deliberately kept separate from US export controls on advanced chips. Keeping those two issues in different rooms reduces the risk that either side uses safety discussions as leverage in the broader technology competition.
What the two sides are actually discussing
Both nations are expected to put forward proposals around voluntary self-monitoring for AI labs, along with frameworks for sharing information about emerging risks.
Chinese state media has been critical of how the US approaches AI governance, framing some American regulatory moves as thinly veiled industrial policy rather than genuine safety measures. That tension will be in the room.
The Bessent-led delegation is an interesting signal in itself. Treasury’s involvement, rather than a purely tech-focused or defense-focused agency, hints at the economic framing both sides are likely to bring.
What the September talks will not resolve
The deeper tension remains between two countries that are simultaneously the world’s leading AI competitors and its two most important AI safety partners. The agenda is narrow by design, and after two years without a dedicated forum, getting back to the table is itself a meaningful development.