The stakes within the case—how a lot the federal government can punish an organization for not taking part in ball—had been obvious from the beginning. Anthropic drew a lot of senior supporters with unlikely bedfellows amongst them, together with former authors of President Trump’s AI coverage.
However Decide Rita Lin’s 43-page opinion means that what is mostly a contract dispute by no means wanted to succeed in such a frenzy. It did so as a result of the federal government disregarded the existing process for a way such disputes are ruled and fueled the hearth with social media posts from officers that may finally contradict the positions it took in court docket. The Pentagon, in different phrases, wished a tradition struggle (on high of the particular struggle in Iran that started hours later).
The federal government used Anthropic’s Claude for a lot of 2025 with out criticism, in keeping with court docket paperwork, whereas the corporate walked a branding tightrope as a safety-focused AI firm that additionally received protection contracts. Protection staff accessing it via Palantir had been required to just accept phrases of a government-specific utilization coverage that Anthropic cofounder Jared Kaplan said “prohibited mass surveillance of Individuals and deadly autonomous warfare” (Kaplan’s declaration to the court docket didn’t embody particulars of the coverage). Solely when the federal government aimed to contract with Anthropic straight did the disagreements start.
What drew the ire of the choose is that when these disagreements grew to become public, that they had extra to do with punishment than simply reducing ties with Anthropic. They usually had a sample: Tweet first, lawyer later.
President Trump’s submit on Reality Social on February 27 referenced “Leftwing nutjobs” at Anthropic and directed each federal company to cease utilizing the corporate’s AI. This was echoed quickly after by Protection Secretary Pete Hegseth, who mentioned he’d direct the Pentagon to label Anthropic a provide chain danger.
Doing so necessitates that the secretary take a particular set of actions, which the choose discovered Hegseth didn’t full. Letters despatched to congressional committees, for instance, mentioned that much less drastic steps had been evaluated and deemed not attainable, with out offering any additional particulars. The federal government additionally mentioned the designation as a provide chain danger was crucial as a result of Anthropic might implement a “kill swap,” however its attorneys later needed to admit it had no proof of that, the choose wrote.
Hegseth’s submit additionally acknowledged that “No contractor, provider, or companion that does enterprise with america navy might conduct any business exercise with Anthropic.” However the authorities’s personal attorneys admitted on Tuesday that the Secretary doesn’t have the ability to do this, and agreed with the choose that the assertion had “completely no authorized impact in any respect.”
The aggressive posts additionally led the choose to additionally conclude that Anthropic was on strong floor in complaining that its First Modification rights had been violated. The federal government, the choose wrote whereas citing the posts, “got down to publicly punish Anthropic for its ‘ideology’ and ‘rhetoric,’ in addition to its ‘vanity’ for being unwilling to compromise these beliefs.”
