Pentagon told Anthropic it was in "near-complete agreement" a week before Trump announced the split

Anthropic has filed two sworn legal statements in a federal court in California, disputing the Pentagon's claim that the company poses an "unacceptable risk to national security." Details of the negotiations have also come to light that call the logic of the entire conflict into question: according to the court filings, just a week before Trump publicly announced the end of cooperation, Defense Department representatives told Anthropic the two sides had "almost reached an agreement."

The conflict between Anthropic and the Pentagon escalated after the Trump administration blocked the company's export license, citing threats to national security. For Anthropic, the stakes are critical: without the license, the company cannot supply its AI products to US allies and large international customers. Anthropic has challenged the decision in court, arguing that it was arbitrary and violates established regulatory procedures.

According to the filings, Anthropic maintains that the government's position rests on technical misunderstandings and on objections that were never raised during months of negotiations. The company offers specific arguments: the Pentagon allegedly misread the architectural features of the Claude models and applied to them security requirements designed for fundamentally different systems.

The key piece of evidence in the case is an internal Defense Department document. According to it, in negotiations a week before the break, Pentagon staff told Anthropic representatives directly that the two sides' positions "almost coincide." This directly contradicts the official rhetoric about "unacceptable risk" and, in Anthropic's view, shows that the decision was political rather than technical.

Anthropic's history with the federal government is mixed. The company has consistently positioned itself as the "safe" alternative player in the AI race. That helped it attract large investments from Amazon and Google and build contacts with the defense establishment. The rupture has hurt Anthropic's reputation as a reliable partner for government bodies, a reputation the company has carefully built over recent years.

Trump said back in February that he considered the relationship with Anthropic "finished," without publicly giving reasons. Even so, Defense Ministry officials continued negotiating with the company for another week after those words, a fact Anthropic now cites as evidence that the government's position is inconsistent.

The case matters beyond Anthropic. It shows how vulnerable AI companies are to politically motivated export-control decisions. As AI development becomes more entangled with national security, the government's ability to block private companies' activities on vague pretexts becomes a real systemic risk for the whole industry.

The next hearing is expected in the coming weeks, and its outcome could set a precedent for the entire sector.