OpenAI defends Pentagon contract amid safety concerns

OpenAI has finalized an agreement to deploy its artificial intelligence models within the Pentagon, the US Department of Defense. The move follows the exclusion of competitor Anthropic. On Friday, negotiations between Anthropic and the Pentagon fell through. Following this, US President Donald Trump directed federal agencies to stop using Anthropic’s technology after a six-month transition period. US Secretary of Defense Pete Hegseth designated the competing AI company as a supply-chain risk.

In response to public scrutiny, OpenAI published a blog post outlining its safety approach. The company highlighted three areas where its models cannot be used:

• Mass domestic surveillance
• Autonomous weapon systems
• High-stakes automated decisions

OpenAI stated that its agreement protects these boundaries through a multi-layered approach. According to the OpenAI blog, the company retains full discretion over its safety stack, deploys via cloud, keeps cleared OpenAI personnel in the loop, and has strong contractual protections. The company added that it did not know why Anthropic could not reach this deal.

However, critics have raised concerns about the contract’s terms. Techdirt writer Mike Masnick reported that the deal permits domestic surveillance because it complies with Executive Order 12333, a US intelligence directive. Masnick noted that this directive is used to capture communications by tapping into lines outside the US, which can include information on US persons.

“Deployment architecture matters more than contract language,” said Katrina Mulligan, OpenAI’s head of national security partnerships. Mulligan argued that limiting deployment to a cloud API ensures models cannot be integrated directly into operational hardware or weapons systems for the Department of War, a term she used for the Department of Defense.

OpenAI CEO Sam Altman acknowledged the intense public backlash, noting on X that the deal was definitely rushed and that the optics do not look good. The controversy coincided with Anthropic’s Claude overtaking OpenAI’s ChatGPT in Apple’s App Store on Saturday. Altman stated that the company wanted to de-escalate tensions between the industry and the defense sector. He noted that if the move fails to de-escalate these relations, OpenAI will continue to be characterized as rushed and uncareful.

Why it matters

OpenAI is defending its new Pentagon contract against public criticism regarding safety and potential for misuse, while simultaneously navigating the fallout of a competitor’s exclusion from government use.