Act in haste, repent at leisure. Defense Secretary Pete Hegseth, take note.
A federal judge on Thursday struck down the Trump administration’s sweeping effort to blacklist Anthropic, ruling that the Pentagon unlawfully retaliated against the artificial intelligence company and exceeded its authorities by labeling the firm a risk to national security.
The Pentagon had good reasons to be frustrated with Anthropic. Its mistake was using a legal bazooka, which transformed a legitimate dispute over the terms of use for software into a sprawling campaign that handed the company an early win in an ongoing legal war.
The confrontation boiled over shortly after the January operation to capture Venezuelan President Nicolás Maduro. An Anthropic executive contacted a counterpart at defense contractor Palantir, asking whether Anthropic’s Claude model had been used. Pentagon officials took the inquiry as suggesting that Anthropic might object to the model’s use in kinetic operations, an interpretation the company disputes.
The episode escalated tensions that had been building for months. Anthropic signed a government contract last year that included restrictions on its models being used in fully autonomous weapons and for mass domestic surveillance. A few months later, the Trump administration began negotiations to loosen those restrictions, insisting the Pentagon must have the authority to use the tools for all lawful purposes. A private company, the government argued, cannot have veto power over decisions made by democratically elected officials. Anthropic was refusing to bend.
The firm’s perhaps casual inquiry into the Venezuela operation incensed the administration. President Donald Trump ordered federal agencies to stop using Anthropic technology, and Hegseth invoked two supply-chain laws aimed at hurting its business.
U.S. District Judge Rita F. Lin rejected that gambit. She held that Hegseth’s invocation of a law aimed at protecting government systems from an “adversary” that might sabotage or subvert them was arbitrary and capricious. She blocked Trump’s government-wide ban and Hegseth’s attempt to prevent other military contractors from doing unrelated business with Anthropic.
But Lin’s ruling does not resolve the matter. The administration may appeal the decision. A second case, under a separate law, remains pending before the U.S. Court of Appeals for the D.C. Circuit. That court has allowed the Pentagon to continue excluding Anthropic while the case proceeds, calling the questions presented “novel and difficult.”
None of this was necessary. Spooked by Anthropic, the Pentagon accelerated efforts to bring competing models onto its classified networks. By May, the military reached agreements with eight other technology companies, reducing dependence on any single entity. When Anthropic subsequently developed Mythos, an advanced model with extraordinary cybersecurity capabilities, the government found reason to work with the company again, even as the Pentagon continued to call Anthropic a supply-chain risk.
The lesson should be obvious. The government is right to be paranoid about private contractors second-guessing its lawful decisions in combat. But lashing out has helped no one.
The post The Pentagon loses a battle in its unnecessary war with Anthropic appeared first on Washington Post.
