Analysis
A federal judge has struck down the Defense Department's designation of Anthropic as a supply-chain risk, the label that had directed federal agencies to stop doing business with the AI lab. U.S. District Judge Rita Lin ruled the designation violated the First Amendment as unlawful retaliation and the Fifth Amendment as a denial of due process, and was in any case arbitrary and capricious, TechCrunch reported Friday morning.
The underlying dispute is about who controls how a purchased model may be used. Anthropic's usage policy restricts fully autonomous weapons targeting and mass domestic surveillance; the Pentagon, under Defense Secretary Pete Hegseth, took the position that a vendor imposing those limits on paid-for systems, and criticizing the government publicly for pushing back, constituted a risk to the supply chain. The Register reported separately that the department's risk assessment cited Claude capabilities the model does not have, a detail that goes directly to Lin's arbitrary-and-capricious finding.
How Anthropic got here
Anthropic was founded in 2021 by siblings Dario and Daniela Amodei with a group of former OpenAI researchers, and has since become the enterprise-facing counterweight to OpenAI, with Claude embedded in AWS Bedrock, Google Cloud and, as of this week, Salesforce's Claudeforce interface. Federal work was a growth line rather than the core business: the department kept pursuing a contract with Anthropic even while the label was in force, and the two sides collaborated on the Mythos model for cybersecurity. Pulse has followed the Anthropic-Pentagon fight since the designation landed.
What the ruling does not settle
A district court order is not the end of the matter. The government can appeal to the Ninth Circuit, and the parallel case in Washington, D.C. remains open. Procurement decisions also have long tails: contracting officers who spent months routing around a vendor do not reverse in a week, and agencies that switched workloads to OpenAI, Google or Microsoft-hosted models are unlikely to switch back on a court order alone. The commercial damage is largely already booked.
The competitive frame
OpenAI, Google, Meta and Palantir have all leaned into defense work with fewer public conditions attached. OpenAI signed a $200 million Defense Department agreement in 2025; Palantir's government business is its foundation. Anthropic's constraint is genuinely differentiated -- and expensive. The ruling means the cost of that position is now measured in lost contracts and legal fees rather than in outright exclusion from the federal market.
The counterweight
It is worth being precise about what the court did and did not find. Lin ruled on the process and the constitutional claims, not on whether Anthropic's usage policy is wise or whether the Pentagon should buy Claude at all. The department retains wide latitude to choose vendors on ordinary procurement grounds. A finding of retaliation constrains the reason it may exclude a vendor, not the outcome.
For other AI vendors, the practical takeaway is narrower than the headline: written usage restrictions are now defensible in court against retaliatory exclusion, but they still cost real revenue in the interim. Watch whether the Defense Department appeals within the 60-day window, and whether any agency restores a paused Anthropic contract before it does.
