OpenAI CEO Sam Altman has acknowledged that the company’s recently announced agreement with the Department of Defense was “definitely rushed,” adding that “the optics don’t look good.”

The deal followed a breakdown in negotiations between the Pentagon and Anthropic. After those talks concluded on Friday, President Donald Trump directed federal agencies to cease using Anthropic’s technology following a six-month transition. Secretary of Defense Pete Hegseth stated he was designating the AI firm as a supply-chain risk. Shortly thereafter, OpenAI announced its own agreement to deploy models in classified environments.

This rapid development raised immediate questions, given that both Anthropic and OpenAI had publicly stated they would not allow their technology to be used for fully autonomous weapons or mass domestic surveillance. Observers wondered whether OpenAI was being truthful about its safeguards and why it succeeded in securing a deal where Anthropic had not.

In response to the growing scrutiny, OpenAI executives took to social media to defend the agreement. The company also published a detailed blog post outlining its approach to safety in national security contexts. The post specified three explicit prohibitions: the technology cannot be used for mass domestic surveillance, autonomous weapon systems, or “high-stakes automated decisions (e.g. systems such as ‘social credit’).”

OpenAI positioned its safeguards as more robust than those of competitors, stating that unlike other AI companies that have “reduced or removed their safety guardrails and relied primarily on usage policies,” its agreement protects its core principles through a “more expansive, multi-layered approach.” This approach includes retaining full control over its safety systems, deploying via cloud API, keeping cleared personnel involved in the process, and establishing strong contractual protections, all in addition to existing U.S. laws.

The blog post concluded by noting, “We don’t know why Anthropic could not reach this deal, and we hope that they and more labs will consider it.”

However, the announcement quickly drew criticism. Mike Masnick of Techdirt argued that the deal “absolutely does allow for domestic surveillance,” pointing to a clause stating that data collection would comply with Executive Order 12333, among other statutes. Masnick described that order as the legal framework that allows the NSA to conduct surveillance on communications outside the U.S., even when they involve American citizens.

OpenAI’s head of national security partnerships, Katrina Mulligan, addressed these concerns in a LinkedIn post. She pushed back against the notion that contract language alone is the primary barrier against misuse, stating, “That’s not how any of this works.” Mulligan emphasized that deployment architecture is more critical than contractual terms, explaining that by limiting deployment to a cloud API, OpenAI can prevent its models from being directly integrated into weapons systems or surveillance hardware.

Amid the controversy, Sam Altman engaged with questions on social media platform X. He conceded the process was rushed and acknowledged the significant backlash, which was so pronounced that Anthropic’s Claude briefly surpassed OpenAI’s ChatGPT in Apple’s App Store rankings. When asked why the company proceeded, Altman explained, “We really wanted to de-escalate things, and we thought the deal on offer was good.”

He framed the decision as a strategic risk, stating, “If we are right and this does lead to a de-escalation between the DoW and the industry, we will look like geniuses, and a company that took on a lot of pain to help the industry. If not, we will continue to be characterized as […] rushed and uncareful.”