Photo Credit: Getty Images
OpenAI says it is revising its recent agreement with the United States Department of Defense after criticism that the rollout appeared rushed and poorly communicated.
On Monday, chief executive Sam Altman acknowledged the company "shouldn't have hurried" the announcement and confirmed new safeguards would be written into the contract, including an explicit ban on using its systems for domestic surveillance of Americans.
The partnership was revealed just hours after federal agencies were instructed to halt the use of rival firm Anthropic's AI tools, following a breakdown in talks over safety commitments. Anthropic had pressed for firm guarantees that its Claude model would not be used for mass surveillance or fully autonomous weapons.
OpenAI said the updated terms would clarify that intelligence bodies, including the National Security Agency, could not deploy its systems without additional contractual changes. The company also stated that the Defense Department understood limits on tracking or monitoring US persons through commercial data sources.
Altman wrote that the issues surrounding military AI were "complex" and required clearer explanation. He said the company had hoped to prevent a more damaging outcome but accepted that the timing "looked opportunistic and sloppy."
The episode highlights growing tension over how artificial intelligence is integrated into defense operations. AI tools are already used to sift satellite imagery, streamline logistics and analyze intelligence at speed. Yet experts warn that large language models can generate inaccurate outputs, raising concerns about reliability in high-stakes environments.
Online reaction to the deal was swift, with some users criticizing OpenAI's involvement in classified defense work and turning instead to competitors. The controversy also renewed debate about how much influence private technology firms should hold in national security matters.
Despite the backlash, OpenAI said it would continue engaging with defense officials while strengthening technical and policy guardrails, arguing that careful collaboration, rather than withdrawal, is the more responsible course.
