OpenAI has secured a Pentagon contract to integrate its artificial intelligence tools into the U.S. military’s classified systems, CEO Sam Altman announced Friday. The deal follows President Donald Trump’s directive to federal agencies to cease using AI from rival firm Anthropic, citing national security risks.
The agreement between OpenAI and the Department of Defense includes specific safety protocols, such as prohibitions on domestic mass surveillance and requirements for human oversight in the use of force, including autonomous weapons. Technical safeguards like full-disk encryption will be deployed, with systems operating exclusively on secure cloud networks. Altman stated the Pentagon demonstrated a “deep respect for safety” and urged the department to extend these terms to all AI vendors, moving away from legal confrontations.
The deal emerges just hours after Trump ordered a six-month halt on government use of Anthropic’s technologies. The president accused Anthropic of attempting to impose restrictive conditions on the Pentagon, writing on Truth Social, “We don’t need them, we don’t want them, and we will no longer deal with them.” This action stemmed from Anthropic’s prior refusal to grant full access to its Claude model for classified networks without binding guarantees against domestic surveillance and fully autonomous targeting. Pentagon officials, including Deputy Defense Secretary Emil Michael, viewed Anthropic’s stance as an effort to dictate terms beyond standard commercial agreements.
Previously, the Department of Defense had been testing Anthropic’s models in secure environments. However, Anthropic’s insistence on ethical guardrails—similar to those now adopted by OpenAI in its military work—led to a breakdown in negotiations. The swift pivot to OpenAI signals a preference for vendors willing to comply without contractual restrictions on potential applications.
The developments underscore a growing divide between AI developers’ internal safety policies and the U.S. government’s push for unimpeded defense technology access. While OpenAI’s deal incorporates its own safety framework, the exclusion of Anthropic highlights tensions over corporate influence on military AI use. This shift may reshape how the Pentagon procures AI, favoring flexibility over pre-negotiated ethical constraints, and sets a precedent for future industry-defense collaborations.
