Trump and Xi Open Door to AI Guardrails in Beijing Summit, But No Substance Yet

President Donald Trump returned from a two-day state visit to Beijing last week saying he and Chinese President Xi Jinping had opened a conversation about coordinating "guardrails" on artificial intelligence, the clearest signal yet that frontier AI has joined chips, defense systems, and trade as a core front in the U.S.-China relationship.

Speaking to reporters aboard Air Force One on the return flight Friday, Trump confirmed the two leaders had discussed "possibly working together" on AI safety measures, though he offered no detail beyond calling them the kind of standard guardrails that come up regularly in policy conversation. When pressed on whether the risks under discussion included biological, nuclear, or cyber threats, he nodded and indicated all three were on the table.

Treasury Secretary Scott Bessent, who has been involved in the early framework discussions, told CNBC that the two AI superpowers were preparing to establish a protocol focused on best practices and preventing non-state actors from gaining access to powerful models. The administration has framed the move as engagement from a position of strength, with Bessent arguing that the United States remains in the lead.

The summit also touched on Nvidia. Trump confirmed that the H200 graphics processing units, which Beijing has not yet approved for import, came up in his meetings. He said he believed something could happen on the issue, though he noted Chinese officials have largely declined to purchase the hardware in favor of developing domestic alternatives. Nvidia CEO Jensen Huang joined the Trump delegation at the last minute, alongside Elon Musk, Tim Cook, and roughly 13 other executives.

Notably absent from the trip were CEOs of leading U.S. AI firms, a gap that critics say reflects how thinly Washington has staffed the technical side of these talks. Former State Department China hand Melanie Hart, now at the Atlantic Council, warned that previous U.S.-China AI safety dialogues under the Biden administration largely failed because Beijing sent foreign ministry officials without deep AI expertise, treating the sessions as intelligence-gathering exercises rather than negotiations.

For now, the Beijing summit has produced what diplomats call a channel, not an agreement. Whether that channel hardens into actual export rules, deployment standards, or testing protocols will likely depend on follow-up working-group sessions over the coming months.