OpenAI Lands Pentagon Deal While Anthropic Declares Its Ethical Principles Absolute

Two very different responses to the same government pressure have defined this week in the AI industry. OpenAI has landed a Pentagon deal it describes as principled. Anthropic has declared its principles absolute and accepted the loss of its government contracts as the price of that declaration. The contrast between these approaches will shape the AI industry’s culture and its relationship with government for the foreseeable future.
The source of the pressure was straightforward. The Pentagon wanted access to Anthropic’s Claude AI system without the company’s ethical restrictions on autonomous weapons and mass surveillance. Anthropic refused to remove those restrictions, offering instead to support all lawful military uses while holding firm on two narrow exceptions. Pentagon officials considered even two exceptions unacceptable, and the administration made its position clear.
President Trump’s intervention converted a commercial negotiation into a political confrontation. His directive ordering all federal agencies to stop using Anthropic technology was sweeping and immediate, and his public condemnation of the company framed its ethical stance as ideological rather than principled. The message to the industry was unambiguous: compliance is expected and non-compliance will have consequences.
Sam Altman chose a path of engagement, announcing a Pentagon deal with assurances that OpenAI’s own ethical commitments — identical to Anthropic’s on the key points — are part of the contract. He sent an internal memo to employees framing the deal as consistent with the company’s values and called publicly for the government to standardize these terms industry-wide. The announcement was accompanied by the closure of a $110 billion funding round at an $840 billion valuation.
Anthropic’s declaration was different in tone and substance. The company stated simply that it had tried in good faith, that no amount of government pressure would change its position on autonomous weapons and mass surveillance, and that its restrictions had never to its knowledge prevented a single legitimate government operation. The company has lost contracts; whether it has lost the argument is a different and longer-running question.