Skip to content
AI
March 3, 20261 min read0 views

Pentagon Blacklists Anthropic Over AI Safety Red Lines, Pivots to OpenAI Deal

TripleG News

TripleG News

18h ago

The U.S. Pentagon has escalated its dispute with AI firm Anthropic by labeling it a 'supply-chain risk to national security,' effectively blacklisting the company from military contracts and suppliers. This rare move against a domestic firm came after Anthropic refused to remove its 'red lines' prohibiting use of its Claude AI for mass domestic surveillance or fully autonomous weapons without human oversight. Defense Secretary Pete Hegseth canceled a $200 million contract awarded last July, accusing Anthropic of seeking veto power over military operations. Anthropic CEO Dario Amodei called the action 'unlawful and politically motivated' and vowed to challenge it in court.

Hours later, OpenAI announced a deal with the Pentagon, embracing an 'all lawful purposes' framework while implementing architectural controls like cloud-only deployment, non-overridable safety stacks, and cleared engineers for oversight. CEO Sam Altman emphasized that these measures align with Anthropic's principles—banning domestic surveillance and ensuring human accountability for lethal force—without relying on contractual prohibitions. Altman noted that existing U.S. laws already cover these areas, positioning OpenAI's approach as more pragmatic than Anthropic's.

The clash underscores deepening tensions between AI ethics and national security needs. Anthropic argues current laws lag behind AI's rapid evolution, risking misuse in surveillance of public data or unpredictable autonomous systems. Pentagon officials counter that vendor-imposed limits undermine operational flexibility, dismissing concerns as ideological. Tech workers have rallied with an open letter urging reversal of the blacklist.

disentangling Claude from classified systems could take months, as seen in its role during recent operations like Operation Epic Fury for intelligence and targeting. Legal experts see potential for Anthropic to prevail, questioning the supply-chain risk label's applicability to contract disputes. The saga may spur broader debates on regulating military AI, with OpenAI pushing for industry-wide safety standards.

Stay Ahead of the Curve

Join 10,000+ tech enthusiasts

Weekly digest · Curated picks · No spam

Related Articles