Anthropic clashes with Pentagon over AI warfare

This is the defining tension of the AI safety era: can labs that built their identity around responsible AI keep those principles intact when the Pentagon comes calling with contracts?

• –Anthropic's usage policies explicitly prohibit Claude from being used to develop weapons or support attacks on critical infrastructure
• –The Pentagon has been aggressively expanding AI procurement, courting every major AI lab as part of DOD modernization
• –Google's 2018 Maven drone controversy set the template — employee revolt, public backlash, eventual retreat — and AI labs are navigating the same minefield now
• –Anthropic's resistance signals that safety-first labs may be genuinely willing to walk away from government revenue, a test of whether their values are real constraints or marketing
• –The outcome shapes whether "safety" remains a meaningful product differentiator or collapses under national security pressure