TL;DR

OpenAI has published a blog post detailing its safeguards within its Pentagon agreement, after CEO Sam Altman admitted the deal was “definitely rushed” and the “optics don’t look good.” The company says its models cannot be used for mass surveillance, autonomous weapons, or social credit-style systems.

A deal under scrutiny

OpenAI moved quickly to sign a military agreement after negotiations between Anthropic and the Pentagon collapsed on Friday. With Anthropic drawing red lines around autonomous weapons and mass surveillance — the same red lines OpenAI claims to hold — critics questioned why one company could reach a deal and the other could not.

In a blog post, OpenAI outlined three areas where its models cannot be used: mass domestic surveillance, autonomous weapon systems, and high-stakes automated decisions such as social credit systems.

The company drew a distinction from competitors, stating that while other AI companies have “reduced or removed their safety guardrails,” OpenAI protects its red lines “through a more expansive, multi-layered approach,” including cloud-only deployment, cleared personnel in the loop, and contractual protections.

Critics push back

The deal has not been without controversy. Techdirt’s Mike Masnick argued the agreement “absolutely does allow for domestic surveillance” because its provisions reference Executive Order 12333, which Masnick described as the mechanism by which “the NSA hides its domestic surveillance by capturing communications by tapping into lines outside the US.”

OpenAI’s head of national security partnerships, Katrina Mulligan, responded that deployment architecture matters more than contract language. “By limiting our deployment to cloud API, we can ensure that our models cannot be integrated directly into weapons systems, sensors, or other operational hardware,” she said.

Altman acknowledges the risk

Altman himself fielded questions on X, where he acknowledged significant backlash — noting that Anthropic’s Claude had overtaken ChatGPT in Apple’s App Store rankings. He framed the deal as an effort to “de-escalate things” between the Department of War and the tech industry.

Looking forward

The deal’s success will likely be measured not by its contract language but by its outcomes. If it leads to de-escalation between government and the AI industry, OpenAI will claim vindication. If not, the company risks being seen as having prioritised access over principle — a charge that may prove difficult to shake.