Moonshot’s new Kimi model has become the latest stress test for Washington’s AI policy, according to MIT Technology Review, because it is free, open-source and competitive enough to put pressure on paid models from OpenAI and Anthropic.
The fight is now visible inside President Donald Trump’s AI circle. MIT Technology Review reported that several current and former Trump advisers attacked major U.S. AI companies in public over the weekend. David Sacks described Anthropic’s models as “lobotomized” and “woke.” Emil Michael, a senior Pentagon official, called OpenAI’s new head of strategic futures a “supreme village idiot.”
The trigger, according to MIT Technology Review, was disagreement over what Washington should do about Kimi, which Moonshot released last week. The model is available at no cost and under an open-source approach, while the leading U.S. systems from OpenAI and Anthropic are commercial products. MIT Technology Review reported that Kimi seems to compete with those systems on capability, creating an awkward problem for U.S. companies selling access to frontier models.
The mechanism is not mysterious. If a capable Chinese model is free, U.S. businesses have a weaker reason to pay OpenAI or Anthropic for similar work. That cuts into the commercial case for American AI labs and complicates a policy pitch that treats those companies as national strategic assets. The politics get messier when the proposed fix is to restrict access to Chinese models, a move that some investors and officials see as self-defeating.
Axios reported that the Trump administration is considering a ban on Chinese AI models after the launch of Kimi K3 renewed calls for tighter restrictions. Fast Company reported that officials are split on the idea. Tech investor Chamath Palihapitiya criticized such a restriction on X, writing: “This would be a terribly self-defeating form of intervention if it were to happen.”
Beijing may be considering its own controls. The Financial Times reported that China is weighing tighter export limits on AI models and chips to stop Western companies from acquiring its technology and startups. Reuters reported that Beijing has discussed possible restrictions with technology firms.
The policy churn is hitting U.S. AI institutions too. Axios reported that Chris Fall, who had led CAISI, the federal AI Safety Institute, since April, has resigned after roughly three months. CNBC reported that no reason was given for his departure.
Anthropic settlement approved
Anthropic also took a major hit in court. Reuters reported that a judge approved the company’s $1.5 billion copyright settlement in a lawsuit brought by plaintiffs who said Anthropic used pirated works to train Claude.
Engadget described the settlement as the largest known copyright payout in history. TechCrunch reported that many authors and creators still do not see the deal as a victory, despite the size of the payment.
The two stories land in the same policy pile: U.S. AI companies want protection from foreign competition, legal certainty over training data and government support for the costs of building large models. Free Chinese releases and copyright lawsuits are both making that harder.
This story draws on original reporting from MIT Technology Review.
