Thursday evening, Judge Rita Lin of the Northern District of California struck down the Trump administration’s designation of Anthropic as a supply-chain risk, ruling that Defense Secretary Pete Hegseth’s February order “constituted unlawful retaliation in violation of the First Amendment” and had “denied the pre-deprivation process required under the Fifth Amendment.” For roughly half a year, the designation had walled the company off from the Pentagon and every federal contractor downstream of it, after Anthropic declined to remove internal guardrails that would’ve permitted the Defense Department to use its Claude models for autonomous weapons and mass surveillance of Americans.

The ruling was blunt about the underlying logic. “The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Lin wrote, a sentence that’ll likely outlive the immediate case.

The politics around it are more tangled than the sentence suggests. In June, President Trump told Axios he no longer viewed Anthropic as a threat. The Pentagon, per TechCrunch, had continued pursuing a contract with the company anyway, collaborating on Anthropic’s new model, Mythos, on cybersecurity work. A supply-chain risk designation in February; a presidential shrug in June; procurement talks running in parallel; a judicial reversal in August. That’s not an oversight regime. That’s four agencies improvising in different directions.

CNN’s Saturday reporting made the architecture explicit. The Center for AI Safety and Innovation, the closest thing to a dedicated federal A.I. body, operates on under $15 million in funding with roughly 30 staff, about one-third the size of its British counterpart. Responsibility for A.I. policy is scattered across the White House Office of the National Cyber Director, Commerce, Treasury, and the Pentagon. One industry source described the situation to CNN as “like early Covid.”

For small businesses that have built customer-outreach pipelines, content workflows, and lead systems on top of a single A.I. vendor, the Anthropic episode isn’t abstract. A February memo took Claude offline for every federal contractor for six months. It’s the second time this year Anthropic’s most powerful models remained offline under an export ban, and it follows secret talks on an IAEA-style A.I. safety body that went nowhere.

Vendor concentration used to be a procurement worry. It’s now a First Amendment one.

Sources