How the latest US China summit could reshape AI

In Beijing this month, Trump and Xi met for the first time in over six months, with AI now sitting alongside trade, Iran, Taiwan and nuclear issues on the agenda. Trump has been explicit that he wants to remind Xi that the United States leads in artificial intelligence, framing AI as a core part of the broader power balance between Washington and Beijing.

At the working level, officials from both sides have started talking about “guardrails” for advanced AI models, with plans for a protocol that aims to stop powerful systems from being misused by non‑state actors. U.S. Treasury Secretary Scott Bessent told CNBC that the goal is to keep innovation moving while building basic safety practices, a conversation sharpened by recent vulnerabilities revealed in Anthropic’s Mythos model.

Export controls remain the hard edge of the rivalry. U.S. chip rules were not the main topic in Beijing, and Washington still does not expect quick approval for Nvidia’s high‑end H200 exports to China, even though roughly ten Chinese firms have been cleared to buy them with added conditions. Beijing continues to push for looser AI‑chip restrictions as part of wider trade talks, arguing that controls on advanced memory and accelerator chips are slowing its domestic AI sector.

The summit also sits against new legislation in Washington that would create a 500 million dollar fund to help allied governments buy U.S. AI models, chips and related tech, explicitly aiming to undercut Chinese AI exports abroad. 

There is now a channel, however thin, to talk about AI safety between the two largest players. At the same time, the race for compute, markets and standards is being formalised in law, export rules and new money.