type: timeline_event
By March 5, 2026, simultaneous pressure from multiple directions had forced a partial reversal of the Pentagon's confrontational posture toward Anthropic. Reports confirmed that Anthropic and the Department of Defense had reopened negotiations, driven by the Pentagon's growing recognition that its operational dependence on Claude during the Iran campaign had severely undermined the credibility of the blacklisting. A senior Pentagon official described the internal moment of realization — when defense leaders became fully aware how deeply Claude was embedded in their targeting systems and what losing access would mean for operational capability — as a "whoa moment."
Simultaneously, a group of prominent defense experts and former national security officials published an open letter to Congress warning that the DoD's supply-chain risk designation against Anthropic set a "dangerous precedent." The letter argued that using the national security designation apparatus to punish a domestic AI company for refusing to remove safety constraints would have lasting chilling effects on the AI industry's willingness to engage with the government, ultimately degrading U.S. defense capabilities rather than enhancing them. The signatories argued that if AI companies learned that insisting on any safety limitations would result in designation as a security threat, the rational corporate response would be to preemptively abandon safety constraints to avoid punishment — a dynamic that would push military AI development in the least safe direction.
Sam Altman, still managing the backlash over OpenAI's own Pentagon deal, publicly echoed this concern, stating he was urging the government to drop Anthropic's supply-chain risk designation. The OpenAI CEO's intervention was notable given that his company had been the immediate commercial beneficiary of Anthropic's blacklisting.
The reopened negotiations, the ongoing lawsuit, the operational dependence on Claude in active combat, and the mounting expert and industry opposition created a situation of acute institutional contradiction for the Pentagon. Having deployed the most severe available regulatory tool against a domestic AI company as leverage in a contract negotiation, the Defense Department found itself simultaneously suing to maintain that designation, relying on the designated company's technology in active warfare, and returning to the negotiating table it had theatrically abandoned. The episode became a case study in the limits of using national security designation powers as a commercial negotiation tactic when the designated company's technology was operationally indispensable.