• Anthropic risks losing a $200 million DoD contract and facing supply chain blacklisting if it does not drop AI safeguards by a Friday 5:01 PM ET deadline.
  • Pentagon CTO Emil Michael publicly rebuked CEO Dario Amodei, calling him a "liar" with a "God-complex" for endangering national security.
  • The dispute highlights escalating tensions between AI ethics and defense needs, with the Pentagon seeking unrestricted military access to AI models.

A Surprising Break in Negotiations

Efforts to restructure Anthropic's AI usage terms with the Pentagon have hit a snag, as the company unexpectedly cut off talks, according to people familiar with the matter. This move comes amid a looming Friday 5:01 PM ET deadline—today, February 27, 2026—set by Defense Secretary Pete Hegseth after a meeting with Anthropic CEO Dario Amodei. Without a deal, Anthropic would be forced into contract termination, supply chain risk designation, and potential invocation of the Defense Production Act to compel compliance.

Chief technology officer Emil Michael, overseeing the negotiations, criticized the breakdown in stark terms. "It's surprising they cut off talks," Michael said, adding that Amodei's stance reflects a "God-complex" that jeopardizes U.S. security. In public remarks, Michael labeled Amodei a "liar," accusing him of prioritizing ideological fears over practical defense applications. Anthropic, an AI safety-focused startup valued at billions, has defended its position, with Amodei emphasizing conscience and risks associated with mass surveillance and fully autonomous weapons. Attempts to reach Anthropic for further comment were unsuccessful.

Financial and Strategic Stakes

The $200 million contract, awarded last summer through Palantir (PLTR), positions Anthropic as the sole AI firm on Pentagon classified networks, deploying its large language model Claude for operations like the recent capture of Venezuela's ex-President Nicolás Maduro. Losing this revenue stream could significantly impact Anthropic's financial performance, though specific details are not disclosed in current reports. Meanwhile, the Pentagon is actively pursuing alternatives, with one AI lab already agreeing to "all lawful purposes" access and two others showing flexibility, according to sources close to the negotiations.

Industry-specific elements are at play: the Pentagon has offered existing U.S. laws and policies as safeguards, plus ethics board invites, but refuses explicit contract language. This reflects broader AI industry shifts toward military applications, with the Pentagon urging four labs for unrestricted use in weapons, intelligence, and combat. If Anthropic fails to meet the deadline, it faces not only contract loss but also supply chain blacklisting at contractors like Boeing (BA) and Lockheed Martin (LMT), potentially reshaping AI-defense partnerships long-term.

Internal and External Repercussions

Anthropic engineers reportedly face internal unrest over military ties, sparking debates on AI ethics versus defense needs. The company recently dropped its core AI safety scaling pledge amid pressure, a move that parallels other labs' negotiations. In the short term, post-deadline actions could include partnership cutoff and legal challenges, with experts noting ideological tensions in AI safety. Michael is already pursuing other AI partnerships, indicating the Pentagon's readiness to move on.

Natural transitions between topics reveal a complex landscape: talks have spanned months since at least mid-February 2026, escalating after Claude's Maduro operation. Historical context includes Anthropic cutting China funding for security, but this standoff tests new limits. For now, all eyes are on the 5:01 PM ET deadline, with Anthropic vowing no concession but anticipating more talks, while the Pentagon prepares for a swift response to secure AI capabilities.

Correction: An earlier version misstated the timeline of threats; they began around February 14, according to reports.