Trump advisors split over response to Chinese open-source AI models

Over the weekend, current and former advisors to President Donald Trump publicly criticized leading US AI companies and questioned how to respond to Chinese open-source models such as Kimi, a recently released model from Chinese AI company Moonshot that appears to rival the intelligence of top US models. The exchanges highlight a widening split among Trump-aligned AI strategists over whether and how the US government should intervene in the AI market.

David Sacks, who served as the president’s AI and crypto advisor until March, publicly labeled Anthropic’s models as 'lobotomized' and 'woke,' and argued that US AI firms seek government protection to eliminate open-source competition. Sacks’ stance—that more open AI is preferable—has reportedly been sidelined within the administration in favor of a view that favors greater government control over AI model releases due to national security concerns.

Dean Ball, a former Trump AI advisor now employed by OpenAI, criticized a new White House review process that vets AI models for security before release, calling it a 'de facto licensing regime for frontier AI.' In response, Emil Michael, a senior Pentagon official and main liaison between the Defense Department and AI companies, dismissed Ball’s warning as a 'supreme village idiot' remark and defended the use of democratic processes over what he described as 'Deep State schemes.'

The debate centers on Kimi and similar Chinese models, which are free to use and impose fewer usage restrictions than their US counterparts. Analysts note that such models are drawing users away from paid US services, complicating the economic and political assumptions underpinning US AI leadership. The infighting suggests no consensus exists within the Trump orbit on whether to counter Chinese AI models through regulation, market dynamics, or diplomatic pressure.

The episode follows a recent decision by the Trump administration to allow Nvidia to sell more advanced chips to China in exchange for a revenue share to the US government, a move that has raised questions about the effectiveness of existing export controls. While the administration alleges some chip smuggling has occurred, it remains unclear what computing infrastructure Chinese firms used to train models like Kimi, with speculation pointing to model distillation techniques that US firms have long complained about.