Trump administration weighs curbs on Chinese AI models as Bessent raises sanctions threat

Bessent said Washington will examine whether Chinese AI models were distilled from American ones, as OpenAI adds veteran financiers to its boards and AI lobbying spending rises.

TRUMP

Summary

The Trump administration is examining whether Chinese AI models were distilled from American systems, with Treasury Secretary Scott Bessent saying Chinese firms could face sanctions if the U.S. determines they are stealing from domestic companies. His comments add to a broader debate over possible U.S. restrictions on Chinese AI as open-weight models from China gain ground against leading offerings from OpenAI and Anthropic. OpenAI is also positioning itself for its next phase of growth, appointing longtime financial executives David Vélez and Robin Vince to its nonprofit and for-profit boards of directors, a move CNBC framed as part of preparations for a potentially massive initial public offering. At the same time, federal disclosures show OpenAI and Anthropic together spent $3.17 million on lobbying in the second quarter, up 23% from the first three months of the year, underscoring how closely major AI developers are engaging with Washington as policy pressure intensifies. The AI story is unfolding alongside a broader shift in U.S. policy and geopolitical risk. U.S. Trade Representative Jamieson Greer said another round of President Donald Trump's tariffs could come soon, suggesting the White House is trying to revive a protectionist trade agenda. In the Middle East, the United States Central Command said it carried out strikes in Iran for an 11th consecutive night, even as reports said regional mediators had presented a 10-day ceasefire proposal.

Terms & Concepts
  • model distillation: A technique for transferring capabilities from a larger model to a smaller one, and a focal point in disputes over whether rival AI systems were improperly derived.
  • open-weight models: AI models whose trained parameters can be shared, allowing others to run and adapt them outside the developer’s own platform.