The Trump administration is facing internal divisions regarding the response to the rapid ascent of China’s advanced AI models, as reported by sources familiar with the ongoing discussions.
The debate contrasts factions within the White House advocating for stricter regulations on Chinese AI technology, which is on the verge of rivaling the most powerful models developed in the U.S., against the Commerce Department’s viewpoint that such regulations may prove ineffective.
This issue has become a significant point of contention for the administration, particularly following the recent unveiling of China’s Kimi K3 model by the Moonshot AI Institute, which competes directly with leading U.S. models from companies like Anthropic and OpenAI.
Officials have indicated that White House staff have been exploring potential measures in recent days to stop Chinese AI laboratories from creating models derived from U.S. technologies—a method referred to as distillation.
The White House is anticipated to take decisive steps after Anthropic Inc. accused Chinese tech giant Alibaba last month of executing what was described as the “largest known distillation attack” to date. Currently, these actions are not expected to manifest as an executive order, according to informed sources.
A press secretary for the White House did not provide comments regarding the reports in this article.
The White House regards distillation attacks with great seriousness. On Wednesday, Treasury Secretary Scott Bessent announced that Moonshot developed the Kimi K3 by applying distillation techniques to Anthropic’s Claude Fable 5 model. He stated on Fox Business that this form of secret distillation could result in sanctions for “intellectual property theft.”
Despite proposals circulating within the White House regarding AI regulations, the Commerce Department, which has its own perspectives, has yet to receive an official request for feedback.
The Department of Commerce manages export controls through the Bureau of Industry and Security, underscoring the growing division within the administration over AI policy.
Lutnick is exploring avenues to incentivize top-tier U.S. AI labs to create their own comprehensive weight models to effectively compete with Chinese advancements. He has engaged in discussions with several AI lab leaders over the past few weeks.
Mr. Lutnick appears to navigate a regulatory middle ground—imposing export restrictions that may keep Anthropic at a disadvantage while adopting a more flexible approach compared to Mr. Cairncross, according to insiders.
Earlier this month, significant questions surrounding AI technologies for governments revolved around Anthropic’s Claude Mythos and Fable 5 models, primarily due to their capabilities in identifying vulnerabilities within government systems and critical infrastructure.
Yet, concerns regarding AI policy are expected to escalate with the emergence of China’s so-called open weight model, which lacks robust safeguards against potential hacking of governmental systems and is accessible for download by anyone.
This is an edition of Hugo Lowell’s innerloop newsletter. Read previous newsletters here.
Source: www.wired.com