OpenAI Signs the Pentagon Deal Anthropic Walked Away From, and Says It Got Better Terms

A week after its rival was barred from military business for refusing to budge, OpenAI announced its own agreement and claimed it had secured every safeguard Anthropic asked for and more. Asked how the terms would be enforced, the company said it could cancel the contract.

OpenAI announced an agreement with the Pentagon this week, a week after Anthropic was declared a national-security supply chain risk for refusing the terms it was offered. OpenAI's claim is pointed: that it secured all of the safety conditions Anthropic had sought — no mass surveillance of Americans, no fully autonomous weapons — plus additional guardrails of its own.

If that is accurate, it complicates the story considerably. The week's easy reading was that one company held to its principles and was punished, while another took the money. OpenAI's version is that the principles were available all along to a company willing to negotiate for them rather than issue ultimatums.

The question is what the conditions are worth, and here OpenAI's own answer is thinner than the claim. Asked how the safeguards would be enforced if the military used the systems in ways the contract forbids, the company said: "We could terminate it if the counterparty violates the terms. We don't expect that to happen." That is the entire enforcement mechanism — the supplier's option to walk away, after the fact, from its largest customer.

It is worth being clear about what nobody can verify from outside. Neither company has published the contract. We have Anthropic's account of the terms it was refused and OpenAI's account of the terms it obtained, and no way to compare them. What can be observed is the outcome: the company that made its limits public and non-negotiable is shut out of government work entirely, and the company that negotiated quietly has the contract. Whatever the merits, that is the lesson other AI companies will draw.