OpenAI has finalized a contract with the U.S. Defense Department allowing its AI systems to operate within the organization's infrastructure, as announced by CEO Sam Altman via a post on X. Altman highlighted two key safety guidelines from OpenAI: bans on large-scale domestic monitoring and requirements for human oversight in decisions involving force, such as those related to self-operating weaponry. He indicated that these guidelines were incorporated into the contract with the department, which the government refers to as the Department of War (DoW), and the department committed to upholding them.
This agreement comes soon after President Donald Trump directed all federal entities to cease utilization of Claude and other offerings from Anthropic. Earlier, U.S. Defense Secretary Pete Hegseth had warned of designating Anthropic as a potential supply chain threat unless it eliminated restrictions in its AI that block applications for widespread surveillance of U.S. citizens and development of completely independent armaments.
The rationale for collaborating with OpenAI remains unspecified, particularly given the comparable restrictions in its systems, though Altman urged the government to extend identical conditions to every AI firm it engages. Jeremy Lewin, serving as Senior Official Under Secretary for Foreign Assistance, Humanitarian Affairs, and Religious Freedom, posted on X that the DoW's agreements cite relevant current laws and incorporate jointly established protective measures. OpenAI and xAI, the latter having previously committed to integrating Grok into the DoW's secure environments, both accepted these provisions. Lewin described it as the identical arrangement presented to Anthropic, which declined it.
Anthropic began its collaborations with the U.S. government in 2024 but resisted pressure from Hegseth. In a recent declaration issued mere hours prior to Altman's disclosure, the company reaffirmed its commitment. 'Threats or penalties from the Department of War cannot alter our opposition to extensive domestic monitoring or self-governing armaments,' the firm stated. 'We intend to contest any supply chain risk classification through legal channels.'
In his X update, Altman noted that OpenAI plans to implement engineering protections to guarantee proper functioning of its AI. He emphasized this aligns with the DoW's expectations and mentioned dispatching technical staff to collaborate with the department on security enhancements. Deployment will be limited to cloud-based infrastructures. According to The New York Times, OpenAI has not integrated with Amazon's cloud services, a platform favored by the government, but recent developments include a new alliance with Amazon to host its models via Amazon Web Services for business users.