The initiative expands access for U.S. military personnel to advanced AI models. They are already using Google’s Gemini and xAI’s Grok.

“We believe that those responsible for defending the country should have access to the best tools. It is also important that the United States and other democratic nations understand how, with proper safeguards in place, AI can help protect people, deter adversaries, and prevent future conflicts,” OpenAI said in its announcement.

The customized version of ChatGPT includes built-in security mechanisms designed to handle sensitive government data.

The chatbot is intended to support routine tasks such as:

  • analyzing and summarizing regulatory documents;

  • preparing materials related to procurement and government contracts;

  • drafting internal reports and checklists;

  • providing administrative and research support.

Risks remain

OpenAI emphasized that all data on the GenAI.mil platform is fully isolated and not used to train the company’s public or commercial models.

However, J.B. Branch of the advocacy group Public Citizen warned of potential risks. In his view, uploading any sensitive information into AI systems inherently increases exposure to external threats.

“Such data should be accessible only to a limited circle of individuals. Even if there is an isolated system authorized solely within military structures, it does not change the fact that classified information is originally intended for a very narrow group,” he said.