A federal judge has delivered a sharp rebuke to the Pentagon, blocking its attempt to punish Anthropic for refusing to remove AI safety guardrails.
The Ruling
US District Judge Rita Lin issued a preliminary injunction on March 26 halting the Pentagon’s designation of Anthropic as a “supply chain risk” — a label typically reserved for foreign adversaries like Huawei and Kaspersky. The ruling also suspends President Trump’s order banning federal agencies from using Anthropic’s Claude AI.
Judge Lin did not mince words. She wrote that “nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government.”
The measures against Anthropic, she found, “appear designed to punish” the company and represent “classic First Amendment retaliation.”
What Anthropic Refused to Do
The dispute stems from Anthropic’s refusal to remove two safety restrictions from its Pentagon contracts:
No autonomous weapons. Anthropic would not allow Claude to be used in systems that select and engage targets without human oversight. The company argued that AI systems are unreliable enough to “hallucinate” — generating false information — making human supervision essential for weapons applications.
No mass surveillance of Americans. Anthropic refused to let Claude be used for large-scale domestic surveillance. AI researchers have warned that aggregating data from over 70 million cameras and financial transaction histories could enable comprehensive population monitoring, creating what they call a “chilling effect on democratic participation.”
These weren’t negotiating positions. CEO Dario Amodei had previously written that large-scale AI-facilitated surveillance should be considered a crime against humanity.
The Pentagon’s Response
When Anthropic wouldn’t budge, the Pentagon escalated. On March 3, Defense Secretary Pete Hegseth designated Anthropic a national security supply chain risk. Trump then ordered all federal agencies to stop using Anthropic’s technology, giving the military six months to phase out systems already in use.
Judge Lin found this retaliation likely illegal. She noted that the Pentagon had previously praised Anthropic and vetted it thoroughly — only turning hostile after the company publicly raised safety concerns.
Why This Matters
This ruling establishes that the government cannot simply crush companies that refuse to build unrestricted surveillance tools. But the underlying threat remains.
The Pentagon wanted Claude without limits — access to an AI system that could process billions of communications, identify patterns of dissent, and flag individuals for investigation at a scale no human workforce could match. That capability still exists. The question is whether any company will be allowed to withhold it.
If Anthropic ultimately loses this fight, the message to every AI company is clear: safety measures exist at the government’s pleasure. Build whatever guardrails you want — but when the Pentagon calls, those guardrails come down.
What Happens Next
The preliminary injunction buys Anthropic time, but the case will continue. The court found the government’s actions “likely” violated the law and the Constitution — but that’s not a final ruling.
Meanwhile, the surveillance infrastructure the Pentagon wants to feed into AI systems keeps growing. Federal agencies already purchase bulk location data, browsing history, and social media activity from commercial data brokers, circumventing warrant requirements by buying what they can’t legally collect.
As Dario Amodei warned, these tools could be used to create “a comprehensive picture of any person’s life — automatically and at massive scale.”
Protect Yourself Now
The outcome of this legal battle will shape AI policy for years. But your personal privacy shouldn’t depend on which side wins in court.
Encrypt your communications. Use Signal for messaging, Proton Mail or Tuta for email. End-to-end encryption means your data can’t be read even if intercepted.
Remove your data from brokers. The commercial data broker pipeline is how agencies get your information without warrants. Opt out. Our free Data Purge tool guides you through the process.
Choose your jurisdiction. Data stored in Switzerland has stronger legal protections than data in the US. Where your data lives matters.
The tools to protect yourself exist today. Use them.