A federal judge in California temporarily blocked the Pentagon from designating Anthropic a supply chain risk and directing government agencies to stop using its AI, according to a March 30, 2026 report by James O’Donnell in MIT Technology Review. The ruling capped a month of escalating conflict that began with social media posts by senior officials and ended with the government conceding in court that several of its stated actions had no legal basis.
- Judge Rita Lin issued a 43-page opinion halting the Pentagon’s supply chain risk designation against Anthropic
- The government admitted in court it had no evidence for its “kill switch” claim, and that Defense Secretary Hegseth lacked authority to bar all DoD contractors from working with Anthropic
- Court documents show the government used Anthropic’s Claude without complaint throughout 2025 via a Palantir deployment, before direct contracting talks revealed policy disagreements
- Anthropic retains a second pending legal challenge; the government had seven days to file an appeal from the ruling
What Happened
U.S. District Judge Rita Lin issued a temporary restraining order blocking the Pentagon from enforcing its supply chain risk designation against Anthropic and from requiring federal agencies to cease using the company’s AI products. The ruling, documented in a 43-page written opinion, came in late March 2026, approximately one month after the dispute first became public. The case drew significant attention because the underlying question — how far the government can go to punish a company for not aligning with its policy preferences — extended well beyond Anthropic alone. MIT Technology Review’s James O’Donnell reported on the decision on March 30, drawing on the written opinion and filings in the case.
Why It Matters
Supply chain risk designations are procedural tools typically applied to companies with ties to foreign adversary governments, not domestic AI contractors with existing government agreements. According to court documents, the government used Anthropic’s Claude throughout much of 2025 without raising objections, with defense employees accessing it through a deployment managed by Palantir. The dispute surfaced only when the two sides began negotiating a direct government contract, at which point disagreements over Anthropic’s usage policies escalated publicly. O’Donnell described Anthropic as having walked a “branding tightrope” as a safety-focused AI company that also pursued defense contracts — a tension that became central to the legal confrontation. Among Anthropic’s unexpected supporters were former authors of President Trump’s own AI policy.
Technical Details
Judge Lin’s opinion identified specific procedural failures in the Pentagon’s handling of the designation. Letters sent to congressional committees stated that less drastic alternatives to the supply chain designation had been considered and rejected, but provided no analysis or supporting details. The government also argued that Anthropic could activate a “kill switch” to deny Claude access to government users — a claim the government’s own lawyers later acknowledged in court they had no evidence to support. Defense Secretary Pete Hegseth had publicly posted that “No contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic,” a statement the government’s lawyers subsequently agreed had “absolutely no legal effect at all.” The judge’s opinion characterized the underlying matter as essentially a contract dispute that was procedurally mishandled and inflamed by social media posts that later contradicted the government’s own court positions.
Who’s Affected
Anthropic cofounder Jared Kaplan submitted a declaration to the court stating that the government-specific usage policy under which Claude was deployed via Palantir “prohibited mass surveillance of Americans and lethal autonomous warfare,” though his declaration did not include the full text of the policy. Defense personnel who had been accessing Claude through Palantir’s platform operated under those terms for the duration of the 2025 deployment period. The ruling has broader relevance for AI companies that hold or seek government contracts while maintaining independent safety and acceptable-use policies that may at times conflict with agency directives.
What’s Next
The government was given seven days from the restraining order to file an appeal. Separately, Anthropic has a second legal challenge against the supply chain risk designation that had not yet been decided as of O’Donnell’s March 30 report. Until both proceedings are resolved, Anthropic remains in a legally uncertain position: the immediate enforcement action is blocked, but the underlying designation has not been formally revoked or withdrawn.
