Judge flags potential retaliation after Pentagon labels Anthropic a security risk

During a San Francisco hearing, US district judge Rita Lin told lawyers she is worried the Department of Defense may have gone too far in labeling Anthropic a supply-chain risk. "It looks like an attempt to cripple Anthropic," she said, adding that the move appears to punish the company for bringing public attention to a contract dispute and could amount to a First Amendment violation.

What Anthropic says

Anthropic has sued the government in two federal cases. The company argues that the security designation was retaliation after Anthropic pushed for limits on how its AI model, Claude, can be used by the military. It is asking the court for a temporary order to pause the designation while the lawsuits proceed. Anthropic says a pause could reassure worried customers and buy time while the legal issues are decided.

What the Defense Department says

The Department of Defense, represented by government attorneys, says it followed the proper procedures and that its officials reasonably determined Anthropic could not be trusted to perform reliably during critical operations. The department has asked the court not to second-guess its national security judgment.

Key exchanges from the hearing

  • Judge Lin noted that deciding whether Anthropic should be a DoD vendor is the secretary of defense 2s responsibility, but she said it is her job to decide whether the department broke the law by going beyond canceling contracts.
  • Lin called it troubling that the designation and the broader directives limiting Claude use by contractors do not appear narrowly tailored to specific national security risks.
  • Government attorney Eric Hamilton said the concern is that Anthropic might deliberately change its software so it would not operate the way the department expects during military use. Hamilton argued that this justified the department 2s assessment of risk.
  • The judge questioned whether the Pentagon considered less severe steps to limit reliance on Anthropic tools before applying a powerful supply-chain-risk label, a tool normally reserved for foreign threats and other hostile actors.

The public actions that escalated the fight

As the dispute intensified last month, Defense Secretary Pete Hegseth posted on social media that, effective immediately, no contractor, supplier, or partner doing business with the US military may conduct commercial activity with Anthropic. At the hearing, the government conceded Hegseth does not have legal authority to bar contractors from using Anthropic for work not related to the Department of Defense. When asked why Hegseth published that message, a government attorney said he did not know.

Industry and transition plans

Defense officials say they are moving to replace Anthropic services over the coming months with alternatives from other AI vendors, including Google, OpenAI, and xAI. The department also says it has put in place measures intended to prevent any tampering during the transition.

Anthropic and its lawyers maintain the company cannot secretly alter its models in ways that would harm government customers. At the hearing, the government attorney acknowledged uncertainty about whether Anthropic could update models without department approval; Anthropic says it cannot.

Next steps

Judge Lin will decide soon whether to issue a temporary pause on the supply-chain-risk designation. To win that pause, Anthropic must show it is likely to succeed on the merits of its claim. A separate appeal at the federal circuit court in Washington, DC, is also expected to be resolved soon, likely without an additional hearing.

Why this matters

The case is about more than one company. It raises broader questions about how government agencies should evaluate technology vendors, how far national security powers extend, and whether companies can publicly press for limits on military applications of their products without facing punitive government action.