The conflict between Anthropic and the Trump administration has widened beyond the Pentagon, turning into a broader confrontation over military use, export controls, and who gets to set the rules for advanced AI systems. What began earlier this year as a dispute over defense applications has now expanded into a federal fight involving the Commerce Department, national security officials, and the company’s latest frontier models.
The latest escalation came after the government ordered Anthropic to suspend access to its newest AI models, Fable 5 and Mythos 5, for all foreign nationals and destinations worldwide. Anthropic complied by disabling the models globally, even though the company said the government’s concerns centered on a narrow potential safeguard bypass in Fable 5 and that the same kinds of software flaws can be found by other publicly available models.
The Commerce Department’s action appears to be unusually aggressive. The order was issued under authorities tied to the 2018 Export Control Reform Act, and export control experts have questioned whether those powers cleanly apply to AI systems typically accessed remotely rather than exported in a conventional physical sense. The government’s stated concern is that Anthropic’s latest models could be exploited by foreign military intelligence, particularly in countries seen as strategic rivals.
That new front in the dispute builds on an already fraught standoff with the Pentagon. Earlier this year, Anthropic refused to allow its models to be used for fully autonomous weapons and the surveillance of Americans, prompting the administration to designate the company a supply-chain risk and direct federal agencies to phase out its products. Anthropic has challenged that action in federal court.
The company’s posture has made it one of the clearest examples of an AI developer trying to enforce its own red lines against the U.S. government. Anthropic has long argued that frontier AI should be deployed with stronger safeguards than many competitors have accepted, especially in areas involving lethal force, domestic surveillance, and cybersecurity. That has placed chief executive Dario Amodei at the center of one of the most visible disputes yet between a major AI company and Washington.
The White House’s broader AI policy has only sharpened the clash. On June 5, President Donald Trump signed an executive order creating a framework for voluntary federal vetting of national security risks in advanced AI systems before release. But Anthropic received the new government directive days later, and the company has said the move runs against the administration’s earlier signals that it would not impose a licensing regime for model reviews.
There are also signs the administration itself is still weighing how hard to press the case. Anthropic’s technical staff have been meeting with Commerce Department officials to negotiate a solution, and Secretary Howard Lutnick has been directly involved in calls with company representatives. That suggests the confrontation, while public and fast-moving, is still being worked through in real time rather than settled.
Outside the company, support for Anthropic’s position has begun to build. More than 100 cybersecurity experts and executives signed a public letter urging the administration to lift the restrictions, arguing that removing leading-edge cyber defense capabilities without a stronger justification could damage U.S. security more than it protects it. The signatories included leaders from major technology companies such as Nvidia and Adobe.
For now, the dispute is no longer just about one contract or one set of models. It has become a larger test of how much leverage the federal government will use over private AI developers, and how much control those companies can retain once their systems are seen as strategically important. With the Pentagon pressure unresolved, export controls now in play, and Anthropic still in court, the feud has moved well beyond a policy disagreement and into a broader fight over power, precedent, and the future governance of frontier AI.
