Anthropic Blacklisted by Trump Administration After Amodei Refuses Pentagon's Final Demand to Strip AI Guardrails
Dario Amodei rejected the Pentagon's 'final offer' on Claude's military use, triggering a supply chain risk designation and government-wide ban. Hours later, OpenAI struck a rival deal with identical safety principles—accepting the same terms Washington called unacceptable from Anthropic.
Overview
The months-long standoff between Anthropic and the U.S. Defense Department reached a definitive and consequential conclusion on February 27, 2026. CEO Dario Amodei refused the Pentagon’s final contract offer, the Trump administration ordered all federal agencies to phase out Anthropic technology, and Defense Secretary Pete Hegseth declared the company a “supply chain risk” — a designation historically reserved for foreign adversaries. Hours later, OpenAI struck its own Pentagon deal, accepting commitments that largely mirrored Anthropic’s contested redlines.
The Final Rejection
Anthropic and the Pentagon had been deadlocked for months over two specific restrictions: that Claude not be used for mass surveillance of American citizens, and that it not power fully autonomous weapons systems making lethal targeting decisions without human authorization. As OPB reported, the Pentagon offered revised contract language on the eve of its self-imposed deadline, but Amodei rejected it as “legalese that would allow those safeguards to be disregarded at will.”
In a public statement, Amodei declared: “We cannot in good conscience accede to their request.” He added that Anthropic had “tried in good faith to reach an agreement with the Department of War, making clear that we support all lawful uses of AI for national security aside from the two narrow exceptions above,” according to ABC News.
Amodei also called out what he characterized as an inherent contradiction in the Pentagon’s posture: simultaneously threatening to brand the company a national security liability while also invoking its essentialness to national security to compel compliance.
The Designation and Its Scope
When the 5:01 p.m. ET deadline passed without agreement, the Pentagon moved swiftly. CBS News reported that Hegseth declared Anthropic a “supply chain risk,” stating: “No contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.” He accused the company of being “sanctimonious” and said “America’s warfighters will never be held hostage by the ideological whims of Big Tech.”
President Trump separately ordered all federal civilian agencies to immediately cease using Anthropic products, according to CNBC. The Pentagon and select defense agencies received a six-month transition period to migrate to alternative systems. Trump also threatened “major civil and criminal consequences” if the company failed to comply.
The supply chain risk designation’s scope is legally disputed. Anthropic cited statute 10 USC 3252, which restricts supply chain risk designations to Pentagon contracting relationships, arguing the order cannot bar military contractors from using Claude for non-defense work, as reported by Axios. The company announced it would challenge the designation in court.
OpenAI Steps In — With the Same Red Lines
The most striking development came hours after Anthropic’s blacklisting, when OpenAI CEO Sam Altman announced his company had reached a deal with the Department of Defense. The terms, as reported by CNBC, included explicit prohibitions on domestic mass surveillance and a requirement for human authorization over lethal force — precisely the two conditions Anthropic had demanded and the Pentagon had refused.
Altman stated: “Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems.” He said the Department of War agreed with these principles, reflecting them in the agreement, and that OpenAI would retain control over how technical safeguards are implemented and which models are deployed.
The Pentagon’s willingness to accept from OpenAI the guardrails it refused to accept from Anthropic drew immediate scrutiny. OpenAI’s arrangement also limits deployment to cloud environments rather than “edge systems” that could include aircraft and drones, and includes forward-deployed OpenAI engineers at the Pentagon to monitor model behavior.
Expert and Industry Reaction
Security policy experts raised questions about both the legal basis and the strategic logic of the supply chain risk designation. DefenseScoop reported that analysts questioned whether the statutory criteria for such a designation had been met, noting the label is typically applied to companies deemed extensions of foreign adversarial governments.
Daniel Castro, Vice President of the Information Technology and Innovation Foundation, warned the move “could send a chilling signal across the broader tech ecosystem.” Tech workers from OpenAI and Google had signed an open letter the night before the deadline expressing support for Amodei’s position.
What It Means
The dispute has exposed a structural tension that neither outcome fully resolves. Anthropic’s refusal, and the administration’s punitive response, establishes that the government is willing to use supply chain designations as leverage against domestic AI companies over compliance disputes — a legal and regulatory tool with no clear precedent in this context. At the same time, OpenAI’s deal demonstrates that the Pentagon’s stated objection was not to safety restrictions per se, but specifically to Anthropic’s version of them.
Anthropic’s contract, valued at up to $200 million, is now terminated. The company has lost access to Pentagon classified networks and faces a government-wide phase-out. Its planned legal challenge, centered on the statutory limits of the supply chain risk designation, represents the next front in a dispute that began with a contract negotiation and has escalated into a constitutional question about the limits of executive power over private technology companies.