Trump orders US agencies off Anthropic as Pentagon labels it a risk

President Donald Trump ordered every US federal agency to stop using Anthropic’s technology on February 27, 2026, and Defense Secretary Pete Hegseth then designated the Claude maker a national security “supply-chain risk”, a label that bars defense contractors from working with the company. The move followed Anthropic’s refusal to drop two limits on military use of its AI: no fully autonomous weapons and no mass domestic surveillance.

This is the first Unicorn API article in this thread. It sets out what was ordered, who is affected, how Anthropic has responded, and what the sources show about where the dispute stands.

What Trump and Hegseth ordered

Trump announced the directive on Truth Social ahead of a 5 p.m. Pentagon deadline. “I am directing EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology. We don’t need it, we don’t want it, and will not do business with them again!” he wrote, adding that there would be a six-month phase-out period for agencies such as the Department of War. The Department of War is a secondary name Trump has given to the Defense Department.

Trump also warned that Anthropic “better get their act together, and be helpful during this phase out period, or I will use the Full Power of the Presidency to make them comply, with major civil and criminal consequences to follow.”

Hegseth followed on X. He wrote that he was directing the department to designate Anthropic a “Supply-Chain Risk to National Security”, adding: “Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.” He said Anthropic would continue to provide its services for no more than six months “to allow for a seamless transition to a better and more patriotic service.”

“America’s warfighters will never be held hostage by the ideological whims of Big Tech. This decision is final.” (Pete Hegseth)

The supply chain risk designation is usually reserved for foreign adversaries. The BBC reported that it would make Anthropic the first US company ever to publicly receive such treatment. A Cloud Security Alliance research note says the label has historically been applied to firms such as Huawei and ZTE, and that it triggers exclusion from defense supply chains and requires existing integrations to be removed.

The two limits at the centre of the dispute

Anthropic’s position came down to two demands: no use of its AI for fully autonomous weapons, meaning AI rather than humans making final battlefield targeting decisions, and no mass domestic surveillance. The Pentagon insisted on “any lawful use” of the company’s tools.

Pentagon spokesperson Sean Parnell said the department had “no interest” in using AI for either task but objected to the guardrails themselves. “We will not let ANY company dictate the terms regarding how we make operational decisions,” he said.

Anthropic CEO Dario Amodei said the company “cannot in good conscience accede” to the request. “In a narrow set of cases, we believe AI can undermine, rather than defend, democratic values. Some uses are also simply outside the bounds of what today’s technology can safely and reliably do,” he wrote. Anthropic told ABC News that the Pentagon’s latest contract language “made virtually no progress” on those points and was paired with “legalese that would allow those safeguards to be disregarded at will.”

The contract in question dates to July 2025, when Anthropic signed a two-year prototype agreement with the Defense Department’s Chief Digital and Artificial Intelligence Office with a ceiling of up to $200 million, executed through a partnership with Palantir. Anthropic was one of four tech firms that won contracts each worth up to $200 million to adapt generative AI for military use. Its tools have been used by the US government and military since 2024, and it was the first advanced AI company to have its tools deployed in agencies doing classified work.

Date (2026)	Event
January	Defense Department begins demanding removal of Anthropic’s two restrictions
February 24	Hegseth issues an ultimatum with a 5:01 p.m. February 27 deadline
February 26	Anthropic publicly declines
February 27	Trump orders agencies to stop using Anthropic; Hegseth announces the risk designation
Early March	Pentagon formally informs Anthropic the designation is effective immediately
March 9	Anthropic files federal lawsuits challenging the designation
September	Defense Department says it has “ceased the use of Anthropic products”

How Anthropic responded and the legal fight

Anthropic called the designation “legally unsound” and said it would “set a dangerous precedent for any American company that negotiates with the government,” vowing to challenge it in court. “No amount of intimidation or punishment from the Department of War will change our position on mass domestic surveillance or fully autonomous weapons,” the company said. It also said it had not received “direct communication” from the Pentagon or the White House on the status of negotiations.

Nearly a week after the initial announcements, the Pentagon said it had “officially informed Anthropic leadership the company and its products are deemed a supply chain risk, effective immediately.” Amodei replied that “we do not believe this action is legally sound, and we see no choice but to challenge it in court.”

Anthropic filed suit seeking to block both the designation and the president’s directive, arguing the administration was illegally retaliating for speech protected by the First Amendment. The filing, made in California, said: “Anthropic’s contracts with the federal government are already being canceled. Current and future contracts with private parties are also in doubt, jeopardizing hundreds of millions of dollars in the near-term.” Microsoft filed an amicus brief supporting Anthropic and urging the court to temporarily block the ban, and legal commentators including Lawfare have assessed the designation as facing significant constitutional and statutory vulnerabilities, according to the Cloud Security Alliance.

The criticism was not limited to Anthropic. Sen. Kirsten Gillibrand called the designation “a dangerous misuse of a tool meant to address adversary-controlled technology,” and Neil Chilson of the Abundance Institute called it “massive overreach.” A group of former defense and national security officials wrote to lawmakers expressing “serious concern.”

Who is affected: government users and contractors

For federal agencies, the instruction is a phase-out. The Pentagon began a six-month phase-out of Claude, and other agencies, including the Treasury Department, said they were discontinuing Anthropic’s products. Claude is reportedly deeply embedded in the federal government, including classified use at the Pentagon, which helps explain the six-month window. In September, the Defense Department said it had “ceased the use of Anthropic products”, including removing them from Palantir’s Maven Smart System.

For contractors, the scope is contested. Hegseth’s post described a ban on “any commercial activity” with Anthropic. Amodei says the Pentagon’s formal notification “plainly applies only to the use of Claude by customers as a direct part of contracts with the Department of War, not all use of Claude by customers who have such contracts.” AFP reported that defense vendors would have to certify they do not use Claude in their Pentagon work. The Associated Press noted that how the department will interpret the scope “is unclear.”

Early responses from large companies show the range:

• Lockheed Martin said it would follow the direction and look to other providers of large language models, expecting “minimal impacts” because it “is not dependent on any single LLM vendor.”
• Microsoft said its lawyers studied the rule and it “can continue to work with Anthropic on non-defense related projects.”

Anthropic says “the vast majority of our customers are unaffected.” It told customers that its priority is “to protect our customers from any disruption caused by these extraordinary events.”

OpenAI’s Pentagon deal and the competitive picture

Late on the same Friday, OpenAI announced a deal to deploy its models on the Pentagon’s classified network. Sam Altman wrote that “two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force,” and that the Pentagon “agrees with these principles... and we put them into our agreement.” He asked the department to offer the same terms to all AI companies.

The Cloud Security Alliance note says OpenAI’s deal keeps the same two restrictions in contract language, but cites critics, including MIT Technology Review and The Intercept, who argue the terms leave room for latitude. It also reports that OpenAI’s head of robotics hardware, Caitlin Kalinowski, resigned over the agreement. Breaking Defense reported that xAI’s Elon Musk and Palantir’s Alex Karp appeared to side with the Pentagon.

The dispute also had a public-facing side effect. PCMag reported that Claude became the top free app on Apple’s App Store, and that Anthropic’s chief product officer said the company was seeing more than a million new customers a day. Amodei apologised for a leaked internal memo in which he criticised Altman.

What it means for builders

For most developers using Claude through Anthropic’s API for commercial work, the sources point to no direct change: Anthropic says the designation covers Claude used as a direct part of Department of War contracts. The picture is different for anyone building for the US military or its contractors.

• Defense-adjacent products need an audit. The Cloud Security Alliance says defense contractors must disclose and report Anthropic integrations in Pentagon-related work, and that obligations cascade through supply chains containing any Anthropic integration.
• Model portability matters. Lockheed Martin’s response, relying on more than one LLM vendor, is the pattern that limited its exposure. The same research note frames this as a new category, “AI vendor geopolitical risk”, where a government action against a provider can force migration across a stack.
• Usage terms are now a commercial variable. The Cloud Security Alliance argues that when one provider is penalised for refusing use requirements, competitors gain an advantage by accepting them, which works against vendor-imposed limits.

Some claims remain unverified. AFP, citing multiple US media reports, said the military used Claude in an attack on Iran after the ban was announced. The sources do not settle the scope question either: Hegseth’s public wording is broad, while Anthropic reads the formal notice narrowly. For background on Anthropic’s financial position during this period, see the desk’s report on Anthropic’s IPO filing.

What to watch next

• The court case. Anthropic’s lawsuits seek to block both the designation and the directive. The sources available to this desk do not show a final ruling.
• A formal executive order. CBS News reported in March that Trump was expected to issue an executive order removing Anthropic’s technology across the executive branch.
• Talks between the parties. Negotiations were reported to have resumed in March, including talks brokered by Anthropic’s investors.
• Contractor guidance. How the Pentagon applies “direct part of” a contract will decide how many Claude users outside government are affected.