History may record this week’s events as the moment the US government definitively asserted its right to use artificial intelligence without ethical constraints imposed by the companies that build it. OpenAI has secured a Pentagon deal that it claims respects its principles; Anthropic has been expelled for holding the same ones.
The story began in the months-long negotiation between Anthropic and the Department of Defense over the terms of AI deployment. Anthropic’s conditions — no use in autonomous weapons, no use in mass surveillance — represented what the company considered a reasonable ethical floor. The Pentagon disagreed, and when negotiations failed, the administration moved against the company decisively.
President Trump’s public condemnation of Anthropic and his directive ordering all federal agencies to cut ties with the company were a pointed demonstration of how far the administration is willing to go to remove ethical constraints from government AI use. The characterization of Anthropic as politically motivated rather than principled was a deliberate attempt to delegitimize its position.
Sam Altman announced OpenAI’s Pentagon deal that same night, insisting the agreement contains the same ethical protections Anthropic had sought. His public statements and internal communications both described autonomous weapons and mass surveillance as fundamental limits that OpenAI will not cross.
Anthropic’s response was concise and defiant: its principles are not negotiable, its restrictions have never harmed a government mission, and political punishment will not change its position. Whether the company’s stand will ultimately be vindicated — commercially, legally, or historically — is a question that will define the next chapter of AI governance in America.