Anthropic fought the US government and won

TL;DR: After almost six months, a California federal court has overturned the government’s blacklisting of Anthropic, calling the move “illegal and baseless.” But, we regret to inform you the Pentagon vs. Anthropic saga still isn’t quite over.

What happened: The court ruled yesterday that the Pentagon designating Anthropic a “supply chain risk” was unlawful. And not just that: It found that the government had violated Anthropic’s First Amendment rights.

“The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Judge Rita Lin wrote in her 59-page ruling. (The decision also says the government ran afoul of the Fifth Amendment right to due process, as Anthropic didn’t get to dispute its blacklisting before it went into effect.)

What started all this, if you’ll recall, was Anthropic’s unwillingness to give the Pentagon free rein on “any lawful use” of its AI. The Pentagon then blacklisted it in early March—banning government agencies (and even its contractors and suppliers) from using Anthropic’s products. The lab sued the government, claiming this freeze-out could cost it billions of dollars in lost business.

Why the court didn’t buy the risk argument: The court concluded that the government didn’t really think Anthropic was a danger. The biggest reason: Right before the blacklisting, the Pentagon was pursuing a contract with Anthropic. And despite accusing the firm of being a national security risk, Defense Secretary Pete Hegseth even suggested invoking the Defense Production Act (which would mark Anthropic as essential to national security) to get it to comply.

Then there was the government’s public hostility toward Anthropic, which focused on its defiance and “betrayal,” not on its models’ potential risks. On the day the blacklisting was announced, President Donald Trump wrote on Truth Social that the US would “NEVER ALLOW A RADICAL LEFT, WOKE COMPANY TO DICTATE HOW OUR GREAT MILITARY FIGHTS AND WINS WARS!” Judge Lin referenced the post in her ruling, among others.

Not a total win: The judge ruled in favor of most of Anthropic’s arguments, but ruled completely against the AI lab on one claim: that Trump didn’t have the authority to order a Claude ban for all federal agencies. (Anthropic simply didn’t make a strong enough case for it, the decision notes.)

What comes next: A separate challenge over the supply chain risk label is still ongoing in DC, but federal agencies must revoke orders enforcing the ban. There’s no statement yet from the government—though it could be prepping an appeal. —WK

Also in AI…

• Basically every tech company we can think of signed a letter yesterday calling for “collective” prep against AI-enabled cyberattacks.
• This headline: “AI industry says Trump plans to tax chips in the ‘single dumbest way imaginable.’”