OpenAI military surveillance autonomous weapons ban is now at the center of the company’s expanded explanation of its new partnership with the U.S. Department of Defense.

OpenAI first announced the agreement on Friday (Feb. 27). A day later, CEO Sam Altman acknowledged in a post on X that the deal was “definitely rushed” and that its “optics don’t look good.”

The announcement came shortly after the Pentagon said it would cut ties with rival AI startup Anthropic.

OpenAI Military Surveillance Autonomous Weapons Ban Explained

In a blog post outlining the agreement, OpenAI detailed three categories where its models cannot be used: mass domestic surveillance, fully autonomous weapons and high-stakes automated decisions such as “social credit” systems.

The OpenAI military surveillance autonomous weapons ban represents two of the same “red lines” that had been sticking points between Anthropic and what OpenAI referred to as the Department of War.

OpenAI said that unlike other AI firms that have reduced or removed safety guardrails in national security deployments, its arrangement protects these boundaries through what it described as a “more expansive, multi-layered approach.”

According to the blog post, OpenAI retains full discretion over its safety stack, deploys its systems via cloud infrastructure, keeps cleared OpenAI personnel in the loop and relies on strong contractual protections, in addition to existing U.S. legal safeguards.

The company added that it was not clear why Anthropic “could not reach this deal,” and expressed hope that other labs would consider similar arrangements.

OpenAI vs Anthropic Pentagon Deal Comparison Intensifies

The OpenAI vs Anthropic Pentagon deal comparison has become unavoidable following the government’s decision to end its relationship with Anthropic.

Last week, the White House directed federal agencies to stop using Anthropic products. President Donald Trump announced that the government would cease working with the company, with use of its Claude models being phased out within six months.

In addition, the government designated Anthropic a supply chain risk, effectively barring companies that do business with the military from working with the startup.

Anthropic responded by stating that the Department of War lacked authority to issue such a designation and said it would challenge the move in court.

In a blog post, Anthropic wrote that no amount of intimidation or punishment would alter its stance on mass domestic surveillance or fully autonomous weapons.

The situation has also intersected with broader geopolitical developments. Published reports over the weekend indicated that Anthropic’s Claude system played a role in U.S. combat operations in Iran, even as the company faces the Anthropic supply chain risk Claude Iran strikes fallout.

As scrutiny continues, the OpenAI military surveillance autonomous weapons ban language appears designed to draw a clear line around permissible uses, while distinguishing OpenAI’s approach from its rival’s failed negotiations.