Sam Altman Admits OpenAI Cannot Control Pentagon's Use of AI Technology

Summary

OpenAI CEO Sam Altman told employees that the company has no control over how the Pentagon uses their AI products in military operations, stating that employees "do not get to make operational decisions" about military strikes. The admission comes amid growing controversy over the Pentagon's demands that AI companies remove safety guardrails from their models to enable broader military applications. AI-enabled systems have reportedly already been used in U.S. military operations to seize Venezuelan leader Nicolás Maduro and in targeting decisions in its war against Iran.

The controversy intensified after Anthropic, OpenAI's rival, refused a Pentagon deal over concerns about domestic mass surveillance and fully autonomous weapons. Defense Secretary Pete Hegseth declared Anthropic a "supply-chain risk" in response—an unprecedented designation for a U.S. company. On the same day, the Pentagon announced a deal with OpenAI, which many viewed as opportunistically replacing Anthropic's Claude chatbot in military applications.

The timing sparked both public and internal employee backlash against OpenAI, with critics suggesting the company crossed ethical lines that Anthropic refused to cross. Altman later admitted the deal announcement was "rushed" and made OpenAI look "opportunistic and sloppy." Anthropic CEO Dario Amodei sharply criticized Altman in an internal memo, calling him "mendacious" and accusing him of producing "safety theater" while praising Trump in "dictator-style." The incident has deepened divisions within the AI industry over military AI applications and the ethical responsibilities of AI companies.

• OpenAI's Pentagon deal announcement—made the same day as Hegseth's threat against Anthropic—triggered internal and external backlash
• Anthropic CEO Dario Amodei accused Altman of "safety theater" and collusion, deepening the rift between the two AI companies