Trump administration AI policy toward China is being shaped by at least seven officials across Commerce, Treasury, the National Cyber Director's office, and the West Wing, with no single coordinating process and sharply different views on how to respond to Chinese open-weight models. The split runs from Commerce Secretary Howard Lutnick's middle-ground stance to Treasury Secretary Scott Bessent's push to add Chinese labs to the Entity List. Former AI czar David Sacks, who left the role in March, still influences policy from the outside via an X account with 1.6 million followers.
The stakes escalate on a fixed calendar. Xi Jinping visits the White House in September, followed by the APEC summit in November, forcing the administration to settle its position on Chinese open-weight AI before year-end. A senior White House official described the internal deliberations to Wired in blunt terms.
“It's not an argument with two sides, it's an argument with ten sides”— Senior White House official, Trump administration
Lutnick has emerged as one of the most active players because Commerce oversees export controls through the Bureau of Industry and Security. He has spoken with leaders at multiple US labs about incentives to release their own open-weight models as a counterweight to China. Last month he used export controls to bring Anthropic to heel on safety commitments, but on other fronts he has taken a lighter touch than the White House hawks.
Key facts
- 01At least seven Trump officials across Commerce, Treasury, and the White House shape AI-China policy with no unified interagency process.
- 02Commerce Secretary Lutnick imposed export controls on Anthropic last month; Treasury Secretary Bessent called Chinese distillation 'IP theft' and floated Entity List sanctions.
- 03Arvind Raman took over the Center for AI Standards and Innovation last week after Chris Fall resigned amid Anthropic negotiations over Fable 5 safeguards.
- 04David Sacks left the AI czar role in March but shapes policy via an X account with 1.6M followers; his July 23 post backing Lutnick drew nearly 500,000 views.
- 05Xi Jinping visits the White House in September ahead of the November APEC summit, forcing decisions on Chinese open-weight models before year-end.
On Thursday Lutnick played down Moonshot AI's Kimi K3, writing on X that his team found it performed worse than leading US frontier models on key benchmarks. His acting deputy at the Center for AI Standards and Innovation, Arvind Raman, took over the agency last week after predecessor Chris Fall abruptly resigned. Fall had spent weeks negotiating with Anthropic's technical teams over jailbreak safeguards for Fable 5, the company's most powerful model, before it was cleared to return online.
National Cyber Director Sean Cairncross has taken a harder line on Chinese labs. He helped assemble Trump's June 2 executive order laying out a framework to assess the most powerful AI models, and he has been empowered inside the White House to develop a national-security-focused response to AI. Cairncross came out of Republican National Committee politics rather than tech, but staff credit him with letting subject-matter experts run without micromanagement. He and chief of staff Lara Smith have focused on stopping Chinese labs from training on outputs from top US models, a practice known as distillation.
Sacks remains a direct line to the president despite no longer holding a formal role. He successfully intervened at the last minute to soften regulatory provisions in the June 2 executive order and has consistently pushed a hands-off approach to Chinese open-weight models. His July 23 post backing Lutnick's Kimi K3 assessment drew nearly 500,000 views.
“Secretary Howard Lutnick is right. The Kimi Panic needs to stop. American frontier models are still ahead. When you factor in what's in the lab, the gap is even larger. As long as we keep releasing, we will stay ahead”— David Sacks, Former Trump AI czar
Bessent has taken the most aggressive posture. As the top Trump official on US-China trade, he called distillation 'IP theft' in an X post last week and threatened Chinese labs with sanctions or Entity List inclusion. He has told White House Chief of Staff Susie Wiles that the US, Europe, Japan, and other Western allies need to accelerate their own open-weight model development or lose the AI race outright.
Bessent has been working Wiles directly, while his deputy Luke Pettit, the Assistant Secretary for Financial Institutions, executes at staff level. Pettit is coordinating with Treasury CIO Sam Corcos, a former DOGE member. Treasury's AI concerns run beyond geopolitics: officials worry a model that inadvertently closed trading positions or accessed sensitive information at banks like Goldman Sachs could destabilize markets.
Wiles sits at the top of the funnel. Lutnick, Bessent, and Cairncross have all taken proposals directly to her, and as with other administration issues, her call is final. The White House frames the resulting policy debate as an innovation contest rather than a regulatory one.
“The United States leads the world in AI innovation, and President Trump will keep it that way. The Trump Administration is doubling down on innovation to widen the gap between America and the rest of the world”— Elizabeth Huston, White House spokesperson
The absence of a unified process cuts both ways. It gives US labs multiple sympathetic ears inside the administration — Anthropic and OpenAI have already leaned on that dynamic to lobby for restrictions on Chinese open-weight competitors. But it also means whichever faction wins the president's attention in a given week can reshape policy without warning, as Sacks did on the June 2 executive order. For companies pricing risk into export-control assumptions or Entity List exposure, that volatility is itself a cost.
The Xi visit in September will force a resolution or expose the fractures publicly. Bessent's Entity List threats and Cairncross's distillation focus point toward escalation; Lutnick and Sacks want to compete on release velocity rather than restriction. Whichever side wins that argument determines whether US AI policy toward China over the next year looks like the semiconductor playbook — hard controls, named entities, blocked flows — or something closer to a market race where Washington bets on speed over walls. For frontier labs, the difference is the difference between a customs regime and a leaderboard.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
