Court rebukes Pentagon, clears path for Anthropic IPO

Judge Rita Lin ruled Thursday that the Pentagon illegally tried to blacklist Anthropic and struck down the Defense Department’s ‘supply‑chain risk’ label.

U.S. District Judge Rita Lin ruled Thursday in California that the Pentagon illegally tried to blacklist Anthropic. The 59‑page decision vacated the Defense Department’s ‘supply‑chain risk’ designation and found Defense Secretary Pete Hegseth engaged in ‘unlawful retaliation’ after the company refused to remove safety guardrails from its Claude model.

The court said Anthropic had set limits that prevent Claude from being used in autonomous lethal weapons and for domestic surveillance. The judge concluded the national security label could not be used to punish the company for those product restrictions.

Anthropic had warned earlier this year the designation threatened billions of dollars in potential enterprise sales and damaged confidence among federal agencies and defense contractors. The court’s order restores the company’s access to federal procurement and contractor channels while appeals proceed.

Anthropic is advancing plans for an initial public offering. The company completed a funding round in May that valued it at $965 billion. Market discussions have placed a possible IPO valuation between $1.5 trillion and $2 trillion. Under the reported IPO structure, some existing shareholders would be allowed to sell portions of their holdings at the offering. Bankers and potential underwriters view the court decision as reducing headline risk ahead of investor presentations planned in September.

Judge Lin framed Anthropic’s safety choices as a form of corporate expression that limits a federal agency’s power to coerce policy changes by threatening exclusion from government business. The Department of Defense must still address a separate procurement proceeding in a Washington appellate court.

Investors, competitors and government buyers are watching the case for its effects on how agencies engage with commercial AI labs that impose usage limits on their products.

Articles by this author