APREX · EST. 2026
When it emerged in February 2026 that the Trump administration had banned federal agencies from using Anthropic's AI models, speculation ran rampant. Was it a technical vulnerability? An AI jailbreak? According to TechCrunch, the answer was no — and the reality is far more politically charged than that.
A $200 million contract — and an abrupt end
In July 2025, the Department of Defense (DOD) entered into a two-year pilot agreement with Anthropic worth up to $200 million. The purpose was to explore the use of the company's Claude models to strengthen U.S. national security. Just months later, it was all over.
In February 2026, President Trump directed all federal agencies to phase out Anthropic's technology. Defense Secretary Pete Hegseth labeled the company a "supply chain risk." Agencies were given six months to wind down their use. The General Services Administration (GSA) simultaneously removed Anthropic from USAi.gov, the government's central platform for AI testing.
The Department of Defense wanted the right to use Claude for "all lawful purposes" — Anthropic put its foot down.
The real fight is over who sets the limits
The core of the dispute was Anthropic's own Acceptable Use Policy (AUP). The company had explicitly prohibited Claude models from being used for two things: mass surveillance of American citizens, and deployment in fully autonomous weapons systems capable of selecting and engaging targets without human oversight.
The Pentagon sought to renegotiate these terms and insisted on using Claude for "all lawful purposes" — without Anthropic having any say over what that specifically entailed. That is where negotiations broke down.
Anthropoc has since emphasized that its reservations were limited to precisely these high-level use cases, not to real-time operational decisions. The company said it was willing to help defense users through the transition.
The federal government's purchasing power as leverage
There is good reason AI companies go to great lengths to retain government contracts. The U.S. federal government spends more than $750 billion on procurement every year, making it one of the world's most influential single buyers.
The Office of Management and Budget (OMB) published new guidelines for federal AI procurement in October 2024, emphasizing risk management and transparency from vendors. At the same time, a draft of new GSA clauses is in the works that would grant the government an "irrevocable license" to use AI systems for "any lawful purpose" — language nearly identical to what the Department of Defense demanded from Anthropic.
A signal to the entire industry
A comprehensive reading of the sequence of events, as TechCrunch analyzes it, points in one direction: the ban on Anthropic is not a technical response to a specific security threat. It is either an expression of political frustration over a private company attempting to set limits on the exercise of state power — or a deliberate punishment for refusing to back down.
Regardless of the motivation, the message to the AI industry is crystal clear: companies seeking federal contracts should be prepared for the government to expect full control over its AI systems — and that ethical guardrails can become a business liability.
How other AI companies choose to navigate this dilemma going forward may well shape the entire global debate over who truly controls artificial intelligence.