Anthropic and the Department of War

The Pentagon has issued Anthropic an ultimatum to modify its existing military contract by Friday at 5pm Eastern, demanding the AI company grant 'unfettered access' to its Claude models. This would require retroactively removing two key safety guardrails from the current agreement: prohibitions against mass domestic surveillance and autonomous kinetic weapons without human oversight. Anthropic, which has been the military's most enthusiastic AI partner and provides the exclusive intelligence engine for Palantir's classified MAVEN Smart System, has strongly signaled it cannot comply with these demands despite previously offering the Pentagon more access than any other entity.

This standoff puts at risk the military's most expensive software license ever purchased—reportedly involving personal endorsement from former President Trump—which currently powers major operations without reported issues. Prediction markets indicate low compliance probability (14%) and significant risk of government retaliation, including Supply Chain Risk designation (16%) or Defense Production Act invocation (23%). The confrontation represents a fundamental clash between military operational demands and AI safety principles, with Anthropic insisting on maintaining its existing contractual red lines while the Pentagon seeks complete operational freedom with advanced AI systems.