Federal Judge Strikes Down Pentagon’s Anthropic Risk Label in First Court Victory for Claude Maker

A federal judge in California has ruled that the Trump administration illegally labeled Anthropic a national-security supply chain risk, handing the Claude maker its first court victory in a fight that could reshape how the government deals with AI companies. The decision says the designation amounted to unconstitutional retaliation and violated Anthropic’s due-process rights.

U.S. District Judge Rita Lin said Thursday evening that the Pentagon’s move against Anthropic was not just improper but “arbitrary and capricious,” finding that officials used national-security language to punish the company after it pushed back on military use cases involving autonomous weapons and mass surveillance.

The ruling is an early but significant win for Anthropic in a broader legal clash with the Defense Department, and it may influence how much leverage federal agencies can use over frontier AI developers that insist on safety limits for government customers.

What did the judge decide?

The court concluded that the government’s supply-chain risk designation could not stand. Judge Lin found that Defense Secretary Pete Hegseth’s labeling of Anthropic as a national-security threat was unlawful retaliation, lacked a rational basis, and deprived the company of constitutionally required process.

In practical terms, the ruling undercuts a sweeping government action that had instructed federal agencies beyond the Pentagon to stop doing business with Anthropic. It also sharply narrows the administration’s argument that it was simply protecting national security by restricting access to Claude models.

The judge said the government’s statements and conduct showed the action was driven by a desire to make an example of Anthropic after the company criticized government policy, rather than by a genuine security assessment.

Lin also drew attention to internal contradictions in the government’s position. On one hand, officials publicly portrayed Anthropic as a risk. On the other, the Department of Defense continued to pursue a contract with the company, discussed using Anthropic’s newer Mythos model for cybersecurity work, and at one point considered using the Defense Production Act in a way that would imply the company was strategically important rather than dangerous.

Why did the Pentagon target Anthropic?

The dispute grew out of Anthropic’s refusal to allow its models to be used for certain military applications. The company drew firm boundaries around uses that could enable fully autonomous weapons systems or mass surveillance of Americans.

Anthropic’s stance reflected a broader strategy it has taken since its founding: emphasizing safety rules and usage restrictions even when they may complicate lucrative government or enterprise deals. In this case, the company argued that the Pentagon could not buy its models and then repurpose them for applications Anthropic considered unacceptable.

Defense officials rejected the premise that they would use the technology for unlawful ends. The Pentagon said its intention was to deploy AI only for legitimate government purposes and suggested Anthropic was trying to dictate how the military could use a tool it had paid for.

The standoff escalated into an unusual public-policy fight over where the boundary lies between AI vendor control and government procurement authority. The administration’s supply-chain risk label represented one of the most aggressive responses yet to a model provider’s safety terms.

How did the court view the national-security argument?

The court said the national-security rationale did not hold up under scrutiny. Judge Lin wrote that the government cannot simply invoke security concerns as a catch-all justification for punishing criticism or disagreement.

She also said Anthropic does not have undisclosed access to the systems it delivers to the Defense Department, weakening the implication that the company could somehow retain hidden control over military use of its models after deployment.

That point matters because supply-chain risk labels are typically associated with hardware sabotage, hidden dependencies, or software vulnerabilities. Here, the court found the government stretched the concept far beyond its ordinary meaning to justify a broader political and contractual dispute.

How did the case get to this point?

Anthropic filed two legal challenges in March, one in California and another in Washington, D.C., after the government imposed the designation and the resulting restrictions. The California case produced Thursday’s ruling, while the separate D.C. lawsuit is still active.

The legal filings marked Anthropic’s attempt to reverse a policy that affected more than just one contract. By extending the label across federal agencies, the administration had effectively signaled that the company should be treated as off-limits to the government as a whole.

That approach raised high-stakes questions for other AI companies as well. If an administration can declare a leading AI provider a supply-chain hazard after a policy disagreement, it could create a powerful precedent for shaping model development, deployment terms, and safety debates by force rather than negotiation.

How does this ruling affect Anthropic’s business and government ties?

The immediate effect is that Anthropic gains legal cover against the specific supply-chain risk label in California, though the broader conflict is not over. The company still has a pending case in Washington, and the government could seek to appeal or pursue other administrative avenues.

For Anthropic, the decision is important not only as a legal win but also as a reputational one. The ruling implies that the company was not acting as a threat to national security simply because it drew hard lines on high-risk military uses.

It also preserves Anthropic’s ability to present itself as a partner that will work with government on constrained, lawful, and security-focused applications. That distinction matters in a market where federal contracts can influence not just revenue but credibility with enterprise buyers and policymakers.

Anthropic said it welcomed the decision and reiterated that it wants to keep working with government on AI systems that support national security while benefiting the public.

What the company wants now

Anthropic’s position is not that it wants to avoid government entirely. Instead, it is trying to define the conditions under which government use is acceptable. The company appears to be arguing that AI firms should be able to refuse applications they view as dangerous without being punished across the federal system.

That argument may resonate with other AI labs, especially those developing frontier models that can be adapted for sensitive uses. It may also appeal to regulators and lawmakers who are increasingly asking how much control vendors should keep after deployment.

What this means for AI policy in Washington

The case arrives at a moment when Washington is still figuring out how to regulate frontier AI while also using it. Federal agencies want access to powerful models for cybersecurity, intelligence analysis, logistics, and software development. At the same time, those agencies are under pressure to prove that AI adoption will not undermine civil liberties or national security.

This ruling suggests courts may be skeptical of government attempts to use procurement powers as a punishment tool when a company objects to military applications. That could give AI companies a stronger hand in negotiations over safety provisions and acceptable-use clauses.

It also raises a broader constitutional issue: whether the government can retaliate against a company’s speech or policy stance by cutting off access to federal business. Judge Lin said the answer, at least on this record, is no.

Why the First and Fifth Amendments matter here

The decision rests on more than contract law. The court tied the government’s actions to the First Amendment, saying Anthropic was effectively targeted for criticizing official policy, and to the Fifth Amendment, saying the company was denied fair process.

That combination is important because it elevates the case from a procurement dispute into a constitutional test of executive power. If the ruling holds, the government may need to show a much clearer factual basis before it can brand an AI vendor a security risk.

For companies working with federal agencies, the implication is straightforward: political disagreement alone should not be enough to trigger a sweeping business ban.

Timeline of the Anthropic-Pentagon dispute

The controversy developed quickly in 2026, moving from policy disagreement to broad federal restrictions and then to court.

Who stands to be affected next?

The immediate parties are Anthropic and the Department of Defense, but the ripple effects could be broader. Other AI companies that sell to government clients may be watching closely to see whether safety restrictions can be defended in court and whether the federal government can retaliate against vendors that refuse certain uses.

Frontier model makers, in particular, are likely to see the case as a signal that a hard line on autonomous weapons and surveillance is not automatically disqualifying for public-sector work. At the same time, the decision may prompt more careful contract drafting and more explicit legal protections for vendor-controlled usage policies.

The ruling could also affect how future administrations approach AI vendors that become politically inconvenient. If the court’s reasoning survives further review, supply-chain risk labels may need a far tighter evidentiary basis than the one the government offered here.

What happens now?

The California ruling does not end the matter. The government can still challenge the decision, and the separate D.C. lawsuit remains unresolved. For now, though, Anthropic has its first courtroom victory in what has become one of the clearest clashes between AI safety policy and federal power.

The case underscores a central tension in the AI era: the same companies that the government wants to rely on for advanced capabilities may also refuse the most sensitive applications. Thursday’s ruling suggests that, at least for now, courts may resist efforts to resolve that tension by branding dissenting vendors as security threats.

Anthropic says it wants to keep collaborating with the government. The question now is whether Washington will be willing to do so without trying to force the company to abandon the safety limits that sparked the dispute in the first place.

Key points at a glance

• A California federal judge ruled the Trump administration’s Anthropic supply-chain risk label was unlawful.
• The court found the move was retaliatory, arbitrary, and a violation of due process.
• The dispute arose after Anthropic resisted military use cases involving autonomous weapons and mass surveillance.
• The decision gives Anthropic a major legal win, but related litigation is still ongoing in Washington, D.C.
• The ruling could influence how federal agencies treat AI vendors that impose safety restrictions on government use.

TechCrunch reported that it has sought comment from the Department of Defense.