Pentagon demands Anthropic strip Claude guardrails

The DoD's hardline stance signals the definitive end of voluntary AI safety pledges when they conflict with national security imperatives.

• –Anthropic faces an existential choice: abandon its foundational safety mission or be locked out of the lucrative defense sector entirely.
• –The ultimatum highlights the tension between AI labs' internal "red lines" and the military's demand for unrestricted model access.
• –OpenAI's rapid acceptance of similar contracts isolates Anthropic, with Dario Amodei publicly criticizing competitors' guardrails as mere "safety theater."
• –The outcome will dictate whether AI developers or the Pentagon ultimately define the ethical boundaries of foundation models in warfare.