Mapping Trump's AI brain trust: seven officials, four departments, ten competing visions for beating China

Ten Competing Visions for Beating China: Inside Trump's Fractured AI Brain Trust

A senior White House official recently described the internal debate over how the US should respond to China's surging open-weight AI models in blunt terms: "It's not an argument with two sides, it's an argument with ten sides" 1.

That fracturing is the story. Behind the public messaging about a unified AI strategy, seven officials scattered across Commerce, Treasury, the White House cyber office, and the orbit of a departed AI czar are pulling in directions that don't reconcile. For anyone building AI tools, exporting chips, or investing in the space, the question isn't who wins the turf war. It's whether ten incompatible visions produce something resembling industrial policy, or whether the result is regulatory whiplash that leaves American companies guessing.

At the Commerce Department, Secretary Howard Lutnick has positioned himself as a swing vote. He oversees export controls through the Bureau of Industry and Security, giving him enormous leverage over which chips and AI systems can leave the country 1. According to people familiar with his thinking, Lutnick has contemplated incentives for leading US labs to develop their own open-weight models to counter China's 1. He publicly downplayed the capabilities of Moonshot AI's Kimi K3 model, saying his team found it performed worse than US frontier models in benchmark tests 1. Yet Lutnick also imposed export controls on Anthropic to force stronger safeguards, suggesting his lighter-touch reputation has limits 1.

His top deputy is Arvind Raman, acting director at the Center for AI Standards and Innovation (CAISI), who stepped into the role after his predecessor, Chris Fall, abruptly resigned following weeks of pressure on Anthropic to strengthen safeguards on its Fable 5 model 1.

At Treasury, Secretary Scott Bessent has taken a markedly harder line. As the top official managing US-China trade relations, Bessent publicly called model distillation "IP theft" and warned that Chinese labs could face sanctions or placement on the Entity List, the government's trade-restriction roster 1. His push has intensified ahead of Chinese president Xi Jinping's White House visit in September and the APEC summit in November 1. According to a person familiar with his thinking, Bessent has argued that the US and its allies need to accelerate their own open-weight model development or risk losing the AI race to China 1. His point person on AI is Luke Pettit, Assistant Secretary for Financial Institutions, who works alongside Treasury's chief information officer and former DOGE member Sam Corcos 1.

National Cyber Director Sean Cairncross represents a third node, pushing for aggressive curbs on Chinese AI labs from inside the White House. He helped assemble Trump's June 2 executive order establishing a framework for assessing powerful AI models, and he and his chief of staff, Lara Smith, have focused on preventing Chinese labs from distilling American models 1. Cairncross is a former Republican National Committee lawyer with no prior tech or AI experience, though staff credit him with taking the national security risks seriously 1.

Working the outside is tech investor David Sacks, who left his role as AI czar in March but maintains a direct line to the president 1. He successfully pushed to water down regulatory provisions in the June 2 executive order and has used his social media following to amplify the argument that American frontier models remain comfortably ahead, publicly backing Lutnick's assessment of Kimi K3 1.

All of these proposals flow through White House Chief of Staff Susie Wiles, who vets every recommendation before it reaches the Oval Office. Bessent, Lutnick, and Cairncross have all taken their positions directly to her, and as WIRED reports, her decision "invariably ends up being final" 1.

For builders and investors, these fault lines translate into three concrete risks. Export controls under Lutnick are unpredictable: he can be hands-off one week and squeeze a company the next. Distillation enforcement, if Bessent prevails, could target any firm whose models have been used to train Chinese competitors, a sprawling category that could engulf much of the frontier lab ecosystem. And Sacks's laissez-faire camp could delay meaningful guardrails long enough for Chinese open-weight models to embed themselves in developer pipelines worldwide, making future restrictions far harder to enforce.

None of these visions requires the others to work. That is the structural problem. When seven officials across four corners of the government each hold a piece of the lever, policy doesn't move in a straight line. It lurches, and the companies trying to plan three years of compute investment, hiring, and international expansion absorb the cost of every jolt.