OpenAI CEO Sam Altman addresses the gathering at the AI Impact Summit, in New Delhi, India, February 19, 2026.
Bhawika Chhabra | Reuters
In a recent address, OpenAI CEO Sam Altman expressed regret over the company’s approach to its new partnership with the U.S. Department of Defense. He acknowledged that the organization “shouldn’t have rushed” into finalizing the agreement and is now looking to make necessary adjustments.
Altman took to X to share insights from an internal memo. He indicated that the company intends to revise the contract to include specific provisions stating that “the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals.”
This announcement follows OpenAI’s recent deal with the Defense Department, which occurred shortly after President Donald Trump instructed federal agencies to cease using competing AI firm’s tools, just hours before planned U.S. military actions in Iran.
Additionally, Altman confirmed that the Pentagon assured OpenAI’s technologies would not be deployed by intelligence agencies like the NSA.
“The technology isn’t fully prepared for numerous applications, and there are still many trade-offs we don’t completely understand concerning safety,” Altman stated. He emphasized that OpenAI will collaborate with the Pentagon to implement technical safeguards.
He further admitted to errors in judgment, remarking that the hasty finalization of the deal on Friday may have come across as “opportunistic and sloppy.” He expressed that their intentions were to mitigate tensions and avert a more severe fallout.
This acknowledgment emerged amid escalating tensions between Anthropic and the government regarding safeguards for its Claude AI systems, a situation that concluded without a resolution. Defense Secretary Pete Hegseth indicated that Anthropic would be classified as a supply-chain threat.
Previously, Anthropic was the first AI lab to integrate its models into the Defense Department’s classified networks following an initial agreement last year.
The company had later sought assurances that its technologies would not be utilized for domestic surveillance in the U.S. or employed in the development of autonomous weapons lacking human oversight.
The conflict intensified when it was disclosed that Anthropic’s Claude was involved in a military operation aimed at capturing Venezuelan President Nicolás Maduro earlier this year, although the company refrained from publicly opposing that particular use.
OpenAI’s agreement with the Defense Department closely followed the breakdown of discussions between Anthropic and the Pentagon. Altman had previously communicated to employees in a Thursday memo that OpenAI shared the same “red lines” as Anthropic. He mentioned in a follow-up post that the Pentagon agreed to their terms.
The reason why the Defense Department agreed to terms with OpenAI but not with Anthropic remains unclear, and government representatives have criticized Anthropic for being overly safety-focused in their approach to AI.
The timing of OpenAI’s deal has led to significant blowback online, with many users reportedly choosing to switch from ChatGPT to Claude on various app stores.
In light of the controversy, Altman stated, “During discussions over the weekend, I emphasized that Anthropic should not be labeled a [supply chain risk] and expressed hope that the [Department of Defense] extends similar terms to them.”
Founded in 2021 by former OpenAI employees, including Dario Amodei, Anthropic has positioned itself as a “safety-first” option in the AI field.
