Anthropic Sues Pentagon After AI Guardrail Dispute Turns Into Federal Blacklist

On Monday, Anthropic filed a lawsuit to block the Pentagon from enforcing a national security blacklist designation that the company says is unconstitutional, unprecedented, and potentially catastrophic for its business. A second lawsuit followed the same day, targeting a broader supply-chain risk designation that could extend the blacklisting across the entire civilian government.

These actions are unprecedented and unlawful. The Constitution does not allow the government to wield its enormous power to punish a company for its protected speech.

-Anthropic

How It Got Here

This didn’t explode overnight. The two sides had been in contentious talks for months over whether Anthropic’s usage policies, specifically its guardrails against autonomous weapons systems and domestic surveillance, were compatible with the Pentagon’s requirement for full flexibility in how it uses AI for “any lawful use”.

Anthropic wouldn’t budge. CEO Dario Amodei has been clear that he isn’t opposed to AI being used in defence contexts in principle, but he believes current AI models simply aren’t reliable enough for fully autonomous weapons. Letting them operate without human oversight, he has argued, would be dangerous. On domestic surveillance, the company drew a harder line, calling it a violation of fundamental rights.

Those positions didn’t sit well with the Pentagon. Defense Secretary Pete Hegseth moved to formally designate Anthropic as a supply-chain risk after the company refused to remove the guardrails. The Pentagon officially informed Anthropic of the designation on March 3. Within days, Trump posted on social media ordering the entire government to stop using Claude. The White House is now reportedly preparing an executive order to formalise that instruction across all federal agencies.

Amodei met with Hegseth in a last-ditch attempt to reach a deal. It didn’t work. Last week a Pentagon official confirmed the two sides were no longer in active talks.

The Financial Damage Is Already Real

Anthropic isn’t just fighting a legal battle; it’s fighting to contain a revenue crisis that is already unfolding.

The company’s finance chief, Krishna Rao, said in court filings that if the government’s actions are allowed to stand, the impact would be “almost impossible to reverse.” Chief Commercial Officer Paul Smith put numbers to that warning: one partner with a multi-million-dollar annual contract has already switched from Claude to a rival AI model, wiping out an anticipated revenue pipeline of more than $100 million. Negotiations with financial institutions worth roughly $180 million combined have been disrupted.

Wedbush analyst Dan Ives summed up the broader risk plainly: enterprises already using Claude may pause deployments while the legal fight plays out, and the reputational damage could linger well beyond any court ruling.

Anthropic executives said the blacklisting could cut the company’s 2026 revenue by multiple billions of dollars. For a startup still building toward profitability, that’s not an abstract threat; it’s existential.

The Industry Is Watching

This fight is bigger than Anthropic. The outcome will shape how every AI company negotiates usage restrictions with the U.S. government going forward and whether private companies or the government have the final word on how AI gets used in military and national security contexts.

A group of 37 researchers and engineers from OpenAI and Google filed an amicus brief in support of Anthropic on Monday, speaking in their personal capacity rather than on behalf of their employers. Among them was Google Chief Scientist Jeff Dean. Their argument: by silencing one lab, the government reduces the entire industry’s ability to openly debate AI’s risks and benefits and discourages the kind of responsible development that ultimately serves everyone’s interests.

The irony here is notable. Anthropic was one of the most aggressive AI companies in courting the U.S. national security apparatus, moving faster than most of its peers in pursuing government contracts. The Defense Department signed agreements worth up to $200 million each with major AI labs, including Anthropic, OpenAI, and Google, in the past year.

OpenAI, for its part, moved quickly to fill the vacuum. The company announced a deal to use its technology in the Pentagon’s network shortly after Hegseth moved against Anthropic. CEO Sam Altman said the Pentagon shared OpenAI’s principles on human oversight of weapon systems and opposing mass surveillance, a pointed contrast to the public framing of Anthropic’s position.

What Comes Next

Anthropic has made clear the lawsuits don’t close the door on negotiation. Company officials said the legal action doesn’t preclude reopening talks with the government and reaching a settlement; they say they don’t want to be in a fight with Washington. But the filing names numerous federal agencies as defendants and asks the court to undo the designation entirely.

The second lawsuit, filed in the U.S. Court of Appeals for the District of Columbia Circuit, targets the broader supply-chain risk designation, the one that could extend the blacklist beyond Pentagon contracts into the entire civilian government. The scope of that designation is still being determined through an interagency review.

However the courts rule, the precedent being set here matters enormously. AI companies have spent years trying to position themselves as responsible partners with governments. Anthropic’s case tests whether that partnership can survive a direct conflict between a company’s stated safety principles and a government’s demand for unrestricted use.