Anthropic CEO Dario Amodei calls OpenAI's Pentagon deal messaging "straight up lies" in staff memo

Anthropic CEO Dario Amodei has launched a direct and unusually public attack on rival OpenAI, accusing the company of dishonest communication regarding its recently announced partnership with the U.S. Department of Defense. In a memo to Anthropic staff that was subsequently leaked, Amodei characterized OpenAI's public statements about the scope and intent of its Pentagon work as "straight up lies," sharply escalating the long-simmering tensions between the two AI safety-focused companies. The conflict centers on OpenAI's announcement that it would lift its ban on military use, with caveats against developing weapons, which it framed as supporting cybersecurity and veteran healthcare projects.

This accusation cuts to the core of the strategic and philosophical divide in the AI industry. While both Anthropic and OpenAI were founded with strong commitments to AI safety, they have diverged in their commercial approaches, particularly regarding government and defense contracts. Amodei's memo suggests Anthropic views OpenAI's nuanced public positioning as a breach of trust and a dangerous precedent. The fallout highlights the intense pressure on AI firms to secure lucrative government deals while maintaining credibility with a workforce and user base deeply concerned about ethical boundaries. This public feud will likely force greater scrutiny on the specific applications of AI models in defense contexts and could influence policy debates.