In July 2025, the Department of Defense awarded contracts of up to $200 million each to four AI companies — Anthropic, Google, OpenAI, and xAI — to “accelerate Department of Defense adoption of advanced AI capabilities to address critical national security challenges.” The contracts were the largest single procurement action for commercial AI in DoD history, and they came alongside deployments across the National Security Council, intelligence agencies, and combatant commands. Anthropic subsequently disclosed that Claude was “reportedly the Department’s most widely deployed and used frontier AI model,” used for intelligence analysis, modeling and simulation, operational planning, cyber operations, and mission-critical applications across the Department of War and other national security agencies. Axios reported in early 2026 that Claude was specifically used in the January 2026 U.S. operation to detain Venezuelan President Nicolás Maduro.

According to court records and Reuters reporting, the tension began in late 2025 when Anthropic inquired about specific DoD use cases. The Pentagon responded by requesting broad contract language allowing use “for all lawful purposes.” Anthropic replied that it was willing to adapt its usage policies for the Pentagon on many points but would not remove its restrictions on fully autonomous weapon systems and mass domestic surveillance of Americans. Anthropic’s public usage policy already prohibited using Claude to incite violence or to design weapons; the contract dispute was about whether the government would honor those existing policy limits as terms of the purchase agreement, not about whether Anthropic would invent new restrictions after the fact.

The Anthropic Pentagon Trump escalation came on February 27, 2026. President Donald Trump directed federal agencies to “IMMEDIATELY CEASE all use of Anthropic’s technology.” Within days, Defense Secretary Pete Hegseth — who now uses “Secretary of War” as a secondary title under Executive Order 14347 signed September 5, 2025 — designated Anthropic a “supply chain risk to national security.” The designation required the U.S. military to stop using Anthropic products and prohibited defense contractors, suppliers, and partners from working with the company, with an up-to-six-month transition period away from Anthropic technology. It was the first time an American company had ever received the label, and it invoked a statute typically reserved for foreign adversaries and companies with demonstrated ties to hostile intelligence services.

Trump’s Truth Social post following the negotiations sharpened the framing. “The United States of America will never allow a radical left, woke company to dictate how our great military fights and wins wars,” he wrote. That framing was echoed by former Trump AI and crypto czar David Sacks, Pentagon Under Secretary Emil Michael, and Hegseth himself in subsequent public statements. Hegseth later posted on X in June that “three months ago, [the Defense Department] kicked @AnthropicAI out of our building — forever.” The ideological framing was not incidental. It was designed to shift the political and press coverage of the Anthropic Pentagon Trump dispute away from the substantive question — whether an AI company can enforce its own use policy as a term of sale — and toward the culture-war question of whether Silicon Valley companies are refusing to serve the American military for political reasons.

Anthropic’s civil complaint in the Northern District of California quantified the damage in specific dollars. Chief Commercial Officer Paul Smith disclosed $150 million in immediate annual recurring revenue loss from cancelled Department of Defense contracts, an additional $100 million lost pipeline from a partner that switched to a rival AI company for a U.S. Food and Drug Administration project, and roughly $180 million in disrupted negotiations with financial institutions worried about the supply chain designation. That totals approximately $430 million in near-term revenue impact from a single dispute.

The financial damage is contextualized by Anthropic’s broader growth. The company crossed $30 billion in annualized run-rate revenue in April 2026, then a $47 billion run rate by late May, at a $965 billion post-money valuation after its $65 billion Series H led by Altimeter, Dragoneer, Greenoaks, and Sequoia. Amazon has committed $13 billion in total, with $5 billion added most recently. But the Pentagon designation put approximately $430 million of confirmed and pipeline revenue at risk, and it complicates Anthropic’s expected 2026 IPO filing, which would need to disclose the ongoing federal litigation and the resolution of the designation.

Anthropic filed suit on March 9, 2026 in the U.S. District Court for the Northern District of California and separately petitioned for review in the D.C. Circuit. On March 26, Judge Lin issued a preliminary injunction blocking enforcement of the supply chain designation. Her ruling described the designation as “Orwellian” and an effort to “cripple” the company. She noted that if the Pentagon’s actual concern was Claude’s reliability in classified operations, the department could simply stop using Claude — it did not have authority to blacklist Anthropic as a national security threat under the statute Hegseth invoked. Lin also emphasized that the statute requires the government to use “the least restrictive means” to address supply-chain concerns, and that the blanket designation of a major American AI company fell well outside that standard.

The D.C. Circuit denied Anthropic’s stay motion on April 8, leaving the company in a split legal position: the N.D. Cal. injunction remains in force, but the designation is nominally active while litigation continues. The court noted that lifting the designation “would force the United States military to prolong its dealings with an unwanted vendor of critical AI services in the middle of a significant ongoing military conflict.” Support then arrived from an unusual quarter. More than 875 researchers and employees at rival OpenAI and Google publicly signed statements supporting Anthropic, arguing that the supply chain risk label could damage the entire American AI industry by allowing the government to punish any lab that maintained its own ethical carve-outs against a single customer. The signatories included Anthropic competitors, not allies. AI critic Gary Marcus separately argued in a widely shared social media post that the designation could push Chinese-born AI researchers currently working at U.S. labs to return to China, undercutting the administration’s own China-competition rationale.

On July 2, 2026, court documents unsealed in the Northern District of California — first reported by the Wall Street Journal and published in full by Gizmodo — revealed a private email exchange between Anthropic CEO Dario Amodei and Emil Michael, the Pentagon’s Under Secretary of Defense for Research and Engineering. The emails cover the final weeks of contract negotiations before the relationship collapsed into litigation, and they show two things that months of public statements had not confirmed. Until the unsealing, both sides had been trading positions in press releases and court filings, but the actual back-and-forth of the negotiation had remained hidden behind confidentiality provisions.

First, the negotiations came “very close” to a deal — the supply chain designation was finalized the day after talks broke down. That timing detail matters because it undercuts the administration’s public framing that Anthropic had never engaged seriously with Pentagon concerns. Second, the core dispute was never about capabilities. Michael described Anthropic’s guardrails as “just not workable” because they carved out categories of use the Pentagon considered essential. Anthropic held the line. Amodei has stated publicly that autonomous weapon systems “may prove critical for our national defense” but should not be reflexively deployed without the technical and ethical infrastructure to keep them within human control. The Anthropic Pentagon Trump fight is now a public record of what happens when an AI lab enforces that line against its largest potential customer, and the unsealed emails will be entered as evidence in the ongoing supply-chain-designation trial that is expected to proceed through the fall.