Dario Amodei’s “We Are Patriots” Stand Is Now a Pentagon Court Fight

Dario Amodei’s February 2026 defense of Anthropic did not end the company’s confrontation with the Pentagon. The dispute has since become an active federal case: a preliminary injunction has restrained important government measures, and the court was still considering the merits after a July 30 hearing.

The central issue remains unchanged. Anthropic was willing to provide Claude for US national-security work, but it would not remove restrictions covering mass domestic surveillance and weapons that select and engage targets without human involvement. What began as a disagreement over those boundaries is now testing how far procurement and national-security powers can reach when an American technology supplier rejects government contract terms.

What Amodei meant by “We are patriots”

The disputed restrictions were narrow rather than a rejection of military AI. Anthropic had already made Claude available for intelligence, cyber, planning and other defense work, and Amodei maintained that the company wanted those deployments to continue under its two exceptions.

The objection to autonomous weapons was partly technical. Amodei’s position was that current models could make targeting errors and lacked the reliability needed for lethal decisions without human supervision. He did not rule out every future autonomous system; he treated reliability, accountability and democratic oversight as unresolved prerequisites.

Mass domestic surveillance raised a different concern. AI could allow the government to process commercially available information about Americans at a scale that older laws and oversight mechanisms did not anticipate. Anthropic therefore treated the capability itself—not only whether a particular data purchase was technically lawful—as requiring congressional scrutiny.

In the CBS News transcript published February 28, Amodei described disagreement with the government as an exercise of First Amendment rights and told President Donald Trump that Anthropic’s leaders were “patriotic Americans.” He ended the answer with the phrase that defined the interview: “We are patriots.”

Why the designation was not a universal ban

The Pentagon’s supply-chain-risk designation was real, but its asserted reach was narrower than a prohibition on every company doing business with Anthropic. The distinction concerned where Claude was being used, not whether an Anthropic customer happened to hold an unrelated defense contract.

Anthropic’s March 5 account of the formal notice states that the company received the Pentagon’s letter on March 4 and understood it to cover Claude when used directly in relevant Pentagon contracts. Under that reading, a contractor could continue using Claude for commercial work or for relationships unrelated to its defense contract.

This qualification does not make the action inconsequential. The supply-chain finding was accompanied by separate efforts to remove Anthropic technology from federal systems and discourage its use across the defense industrial base. Those measures involved overlapping institutions but did not all derive from the same legal authority, which is why treating them as one seamless government-wide ban obscures the court dispute.

Anthropic also offered continued model access and engineering assistance during any transition away from Claude. That offer reinforced the distinction at the heart of Amodei’s interview: the company was declining two categories of use, not seeking to interrupt active national-security work or claim authority over military operations.

The court fight changed the practical status

Anthropic challenged the government’s response under constitutional and administrative law. Its case treats the designation and related directives as retaliation for protected disagreement and as an improper use of a statute intended to protect defense supply chains. The government has defended its authority to decide which suppliers can be trusted in sensitive operations and to demand technology available for all lawful purposes.

A California federal court granted Anthropic preliminary protection in March. That relief temporarily restrained enforcement of major parts of the government’s response, but it was not a final judgment: a preliminary injunction preserves the position during litigation and reflects an initial evaluation of the claims.

At the next major stage, both sides sought summary judgment. Axios’s account of the July 30 hearing records Judge Rita Lin’s concern that the government had not supplied stronger evidence and her question about whether Anthropic could alter or remotely disable a model after delivery. The same account notes that the judge did not set a deadline for her written decision.

The hearing exposed a technical point with legal consequences. If Anthropic cannot modify a model once it is deployed inside a classified environment, the government’s concern about the company interrupting that model becomes harder to separate from negotiations over future versions, support and access. Those are related operational questions, but they are not identical security risks.

What has—and has not—been decided

The preliminary injunction weakened the immediate effect of the Pentagon’s actions, but it did not erase the designation through a final merits ruling. Nor did it require the Pentagon to purchase Claude. The military’s freedom to select another supplier is distinct from the government’s ability to impose broader penalties on Anthropic and its customers.

Amodei’s underlying position also remains more limited than the slogans used by either side. Anthropic has not claimed that a private laboratory should permanently set national policy, and its chief executive identified Congress as the appropriate institution for durable rules. The company’s interim position is that it can refuse contract terms while lawmakers consider questions created by fast-changing AI capabilities.

The unresolved case therefore reaches beyond one model or procurement negotiation. It concerns whether refusal to support particular lawful-but-contested uses can itself justify treating a domestic supplier as a national-security risk, and whether the government may extend the consequences beyond the contract it was negotiating.

For now, the most accurate status is narrower than either “Anthropic won” or “Anthropic was banned.” The Pentagon issued the designation, Anthropic obtained temporary judicial protection, and the merits remained undecided after the July hearing. Amodei’s “We are patriots” statement has become the public shorthand for that continuing argument rather than the final word on it.

Also read:

• How’s Allbirds Doing After Its Wild AI Pivot? Now It’s Smartbird — and the Sneakers Are Officially History
• Part 9: The Great Exit – How Insiders, VCs, Governments, and Even Bitcoin’s Biggest Corporate Holder Are Dumping Right Now
• Are Contractors Liable After Onsite Accidents?