type: timeline_event
On February 26, 2026, Anthropic CEO Dario Amodei issued a formal rejection of the Pentagon's ultimatum, publicly declaring "We cannot in good conscience accede to their request" as the Defense Department's deadline approached. In an exclusive CBS News interview, Amodei stated that Anthropic was in agreement with 98 to 99 percent of the military's proposed use cases for its Claude AI models, but maintained two absolute limits—no use of AI for mass domestic surveillance of Americans, and no use of AI to direct fully autonomous weapons systems that could fire without human authorization.
Amodei framed these restrictions not as anti-military positions but as principled limits grounded in the state of current AI technology: "Use cases like mass surveillance and autonomous weapons are outside the bounds of what today's technology can safely and reliably do." He told CBS that Anthropic's threats "do not change our position" and emphasized that the company considered itself composed of "patriotic Americans" committed to national defense but unwilling to endorse applications that the technology was not yet ready to support safely.
The Pentagon had made a final offer insisting that Anthropic agree to allow Claude to be used "for all lawful purposes" without enumerated restrictions. Anthropic rejected this framing on the grounds that "lawful" is a moving target—many surveillance activities that are technically legal under current statutes could amount to mass monitoring of American citizens without warrants. Anthropic had specifically sought written guarantees that Claude would not be used for those applications.
The rejection was notable because Anthropic had, under the same pressure in the preceding days, already dropped core clauses from its Responsible Scaling Policy. That capitulation made the subsequent refusal on the two core red lines more consequential: the company had conceded on internal policy documents while holding firm on contract language. Defense Secretary Pete Hegseth's 5:01 p.m. deadline on February 27 was now hours away, and Anthropic had made clear it would not comply. The stage was set for the most aggressive use of government contracting power against a domestic AI company in American history.