Updated Feb. 27, 2026 — Washington
President Trump on Friday directed all federal agencies to immediately stop using artificial intelligence tools from Anthropic, escalating a dispute over how the company’s Claude model can be used by the U.S. military. In a Truth Social post, Trump said federal agencies should “IMMEDIATELY CEASE all use of Anthropic’s technology,” adding that some departments, including the Defense Department, would have six months to transition away from Anthropic products and warning of potential civil and criminal consequences if the company does not cooperate.
Within roughly 90 minutes of the president’s post, Defense Secretary Pete Hegseth labeled Anthropic a supply-chain risk, a designation that takes effect immediately and bars any contractor, supplier or partner that works with the U.S. military from commercial activity with the company. Hegseth said the designation allows Anthropic up to six months to finish providing services to the Pentagon to avoid disruption while the department transitions to another provider, and defended the move as a way to prevent warfighters from being “held hostage by the ideological whims of Big Tech.”
Anthropic’s CEO, Dario Amodei, called the government actions retaliatory and punitive in a CBS News interview Friday evening. He said the company had not received formal notification of the supply-chain designation and plans to challenge the decision in court, while remaining open to negotiating an agreement that respects the company’s stated “red lines.” Anthropic argued the Pentagon’s action is legally questionable and risks setting a dangerous precedent, saying the defense official lacks the authority to broadly ban military contractors from doing business with the company outside of Pentagon work.
The conflict centers on the conditions Anthropic built into its Claude model and the Defense Department’s insistence that it be able to use the model “for all lawful purposes.” The Pentagon gave Anthropic a Friday deadline to remove certain safeguards. Anthropic had sought explicit limits, including bans on using Claude for domestic mass surveillance and guarantees that Claude would never make final targeting decisions without human oversight. A person familiar with the matter told CBS News that Claude is prone to “hallucinations” and, as currently configured, is not reliable enough to make potentially lethal decisions without human judgment.
Anthropic won a roughly $200 million defense contract last July to develop AI capabilities for national security and, via a partnership with Palantir, is currently the only company with its model running on the Pentagon’s classified networks. A senior Pentagon official said Grok, the model from Elon Musk’s xAI, could also be used in classified settings.
OpenAI CEO Sam Altman said late Friday that his company reached an agreement with the Department of Defense to deploy its models on the department’s classified network. Altman said the arrangement includes two key safety principles — prohibitions on domestic mass surveillance and human responsibility for the use of force, including in autonomous weapon systems — and called on the Defense Department to extend the same terms to other AI firms.
Tensions escalated this week as the Pentagon and Anthropic negotiated contract language. Pentagon chief technology officer Emil Michael told CBS News the military made concessions, proposing wording that references federal surveillance laws and longstanding Pentagon policies on autonomous weapons. Anthropic argued the revised language still failed to prevent Claude’s use for mass surveillance or fully autonomous weapons and contained legal phrasing that could be used to evade promised safeguards. Amodei said the company preferred to continue serving the Defense Department under the two safeguards and, if the Pentagon chose to offboard Anthropic, the company would try to ensure a smooth transition to avoid operational disruption.
The dispute turned personal on social media, where Michael publicly attacked Amodei, calling him a “liar” and accusing him of having a “God-complex.”
The president’s directive drew concern from both parties in Congress. Sen. Mark Warner (D-Va.), vice chair of the Senate Intelligence Committee, accused Trump and Hegseth of trying to force Anthropic to deploy “AI-driven weapons without safeguards” and said the move should “scare the hell out of all of us,” questioning whether the decision was driven by politics rather than national security analysis.
The situation is likely to trigger legal challenges and further negotiations over how civilian and military priorities, safety restrictions and classified requirements are balanced as AI systems are integrated into national security work.
Reporting contributed by Caitlin Yilek, Jennifer Jacobs and Jo Ling Kent.