Anthropic Was Illegally Blacklisted by the Trump Administration, Court Rules

On August 28, 2026, a federal judge in California delivered a 59-page ruling striking down the Pentagon's designation of Anthropic as a "supply chain risk" — a national-security classification previously reserved for foreign adversaries — finding it to be unconstitutional retaliation against a company that refused to remove safety restrictions from its AI contracts. The ruling takes effect immediately, though the government retains the right to appeal.

The Claim

The dispute traces back to early 2026, when Defense Secretary Pete Hegseth renegotiated the military's AI vendor contracts across the industry, seeking to remove existing restrictions and open the door to what the Pentagon described as effectively any lawful use of AI tools. The renegotiation was sweeping. Most major players in the AI industry — OpenAI, Google, Microsoft, and SpaceX among them — signed revised agreements under the new terms.

Anthropic did not. The company drew two specific lines: its Claude AI models would not be used for mass surveillance of American citizens, and they would not be used to power lethal autonomous weapons systems. Those positions, which Anthropic had defended publicly, became the flashpoint. In February 2026, the Trump administration responded by designating Anthropic a "supply chain risk" under national-security authority — a designation The Guardian confirms had never previously been applied to a domestic American AI company — and replaced it on relevant contracts with seven other AI labs.

Anthropic filed suit in March 2026 in the Northern District of California. Judge Rita Lin issued a temporary suspension of the supply chain risk designation almost immediately after the complaint was filed. On August 28, she made that ruling permanent, finding the designation both "arbitrary and capricious" under administrative law standards and unlawful First Amendment retaliation. In her decision, Lin stated that an empty invocation of national security does not give the government a blank check to punish and retaliate against critics — a formulation that cuts to the constitutional core of the case.

What We See

The First Amendment angle is the most legally significant element of Lin's ruling, and it is worth understanding precisely what it establishes. The Pentagon did not simply choose a different vendor on neutral procurement grounds. The court found that the government specifically targeted Anthropic because of public statements the company had made criticizing military AI policy. That distinction converts what looked like a contracting dispute into viewpoint-based government retaliation — a category courts have long treated as a constitutional violation, regardless of how the national security justification is framed.

Multiple reports confirm that the "supply chain risk" designation had historically been applied to foreign entities suspected of posing security risks to American systems — not to domestic companies objecting to contract terms on ethical grounds. Our read is that this is the more durable portion of the ruling: it sets a meaningful limit on the government's ability to weaponize national-security classifications against domestic vendors who publicly criticize policy or decline specific contract terms. That principle extends well beyond AI procurement, and it is the part of this decision most likely to be tested on appeal.

For developers and companies that have built products on top of Anthropic's Claude API, the ruling closes a vendor-stability risk that had been open since February. The practical concern during the blacklist period was not simply Anthropic's revenue — it was whether the combined financial and reputational pressure, which Anthropic executives said could cost the company billions in lost business, would eventually force the company to soften its stated safety constraints to survive. A company facing contract exclusion and reputational damage has real incentive to compromise. That leverage is now gone. Anthropic's two hard lines — no mass surveillance of Americans, no lethal autonomous AI — have been tested in federal court and held.

A separate source notes the ruling's geographic precision: it bars the specific federal agencies named in the lawsuit from enforcing the Trump order or the Hegseth supply chain designation. An Anthropic spokesperson confirmed the company welcomes the court's decision.

Where It Falls Short

What neither source fully addresses is the industry asymmetry the ruling leaves untouched. Seven AI labs signed contracts under the terms Anthropic refused. The court decision reverses Anthropic's exclusion, but it does not impose any restrictions on how the tools from those other labs can be deployed by the Pentagon. The ruling is a meaningful legal win for one company in a specific retaliation dispute. It is not a ruling on whether AI should be used for surveillance or autonomous weapons — that policy question remains entirely open, and the contracts enabling such use remain in effect elsewhere.

The appeal question is also genuinely unresolved. The government has strong institutional incentive to challenge any ruling that constrains its ability to invoke national-security authority in vendor disputes. Appellate review of executive-branch national-security claims is notoriously difficult terrain, and Lin's First Amendment retaliation theory, while well-reasoned, will face scrutiny at the circuit level. Developers whose products depend on Anthropic's models have legitimate cause for confidence today. Treating this as a fully settled matter before any appeal concludes would be premature.

Sources