Ro Khanna Urges US-China AI Treaty Ahead of Trump Tech Summit

Rep. Ro Khanna (D-CA) is calling for a treaty between the United States and China aimed at preventing artificial intelligence from causing global harm, according to letters he shared with The Verge. The push comes as President Donald Trump prepares to convene tech and AI executives in Washington, putting the question of AI governance directly in front of the industry leaders whose systems would be most affected. Khanna's proposal would make the two countries — currently the world's dominant AI powers — formal partners in setting limits on how the technology is built and deployed.

The timing is deliberate. Trump's meeting with CEOs is expected to cover competitiveness, chip policy, and the widening race to build more capable models, a race in which Washington and Beijing have largely treated each other as rivals rather than collaborators. Khanna's letters, addressed to figures in both governments and the private sector, argue that the current trajectory — each side accelerating while the other watches — leaves no mechanism for catching a catastrophic failure before it spreads across borders. A treaty would create that mechanism, at least on paper.

That is where the difficulty starts. Getting the White House and Beijing to sign onto a shared set of AI rules would require both governments to accept outside scrutiny of programs they treat as matters of national security. Export controls on advanced semiconductors remain a flashpoint, and any agreement touching model development or military AI would run into the same distrust that has stalled previous arms-control-style talks. Khanna's own party does not speak with one voice on China policy, and the Republican-controlled Congress is unlikely to embrace a treaty framework advanced by a Democratic representative.

The letters also land at a moment when the practical gap between the two countries is narrowing in some areas and widening in others. Chinese labs have released capable open-weight models that compete with US systems on benchmarks, while American firms retain an edge in the most advanced chips and in capital spending. That mix of parity and asymmetry is exactly what makes a bilateral deal hard to design: the two sides do not agree on what they would be giving up, or what they would gain.

Khanna's effort may ultimately function less as a near-term legislative vehicle than as a marker of where the debate is heading. If Trump's meeting produces any signal on AI safety cooperation, it would give the idea momentum; if it produces only a competitiveness agenda, Khanna's treaty push becomes another proposal waiting for a crisis to make it urgent. The open question is whether either government sees enough shared risk to accept shared limits before an incident forces the issue.