Chinese AI models spark an internal war within Trump's circle

By Thomas Macaulay · 07.21.26.

Last weekend, several of President Donald Trump's current and former AI advisers publicly traded insults over how to respond to the rise of Chinese models. David Sacks called Anthropic's models "lobotomized" and "woke," while Emil Michael, a senior Pentagon official, called OpenAI's new head of strategic futures "the supreme village idiot."

The dispute was set off by Kimi, a free, open-source model released last week by the Chinese company Moonshot. According to the article, Kimi appears to rival the intelligence of OpenAI's and Anthropic's models, which, unlike Kimi, are not free.

The underlying issue is economic and political: every time China releases a powerful, free model, U.S. companies have fewer reasons to pay for Anthropic's or OpenAI's models. This creates a dilemma for the Trump administration, which sees its rhetoric of American AI leadership clash with the commercial appeal of open-source Chinese alternatives, and it has ended up dividing the president's top AI strategists.

The original article, bylined by James O'Donnell, is part of The Algorithm, MIT Technology Review's weekly AI newsletter.

The same "The Download" edition also includes a section of featured links (not developed in the body of the email) on related topics: the court's approval of Anthropic's $1.5 billion settlement over copyright lawsuits; the possibility that the Trump administration might ban Chinese AI models after the launch of Kimi K3; China's plans to tighten export controls on its AI models and chips; the resignation, after just three months on the job, of the Trump administration's head of AI safety (CAISI); and a report on a new Google chip, dubbed "Frozen V2," aimed at running Gemini models more efficiently.