OpenAI Took the Pentagon Deal — Then Caved to the Same Terms Anthropic Died For

OpenAI secured a $200M contract with the Pentagon, but the initial agreement lacked the explicit bans on surveillance and use against US persons that had been a sticking point for competitor Anthropic. The omission triggered a swift internal revolt, with approximately 900 tech workers—including about 100 OpenAI employees—signing an open letter in solidarity with Anthropic's original ethical stance. Facing this pressure, OpenAI CEO Sam Altman publicly criticized the deal's drafting and moved to amend it, adding the very language that had led to Anthropic's rejection and subsequent blacklisting by the Department of Defense.

The amended contract now explicitly prohibits the use of OpenAI's AI for surveillance, targeting US persons, or utilizing commercially acquired data like location and browsing history. Altman emphasized a personal commitment to ethical boundaries, stating he would rather face jail than follow an unconstitutional order. The critical twist is that the Pentagon confirmed this revised language is 'a compromise that Anthropic was offered, and rejected.' This leaves OpenAI holding the lucrative contract under Anthropic's principles, while Anthropic itself remains excluded from Pentagon work, creating a stark divide in corporate strategy and raising unresolved questions about the internal calculus behind Anthropic's refusal.