Pentagon integrates two customized versions of ChatGPT and Grok into its secured defense portal

The U.S. Department of Defense launched two customized versions of OpenAI’s ChatGPT and XAI’s Grok to provide generative AI tools to roughly three million civilian and military employees through its unified portal. The move makes the ChatGPT Meil and Grok for Government versions available inside the secured central GenAI.mil portal that was launched last year and initially offered Google’s Gemini model, aiming to enable department personnel to use the latest leading commercial models without routing sensitive government data through public consumer channels.

This defensive environment exempts official military accounts from data-collection mechanisms that are difficult to avoid in consumer-oriented commercial products. Department of Defense data showed broad uptake across its sectors, with the portal attracting more than 1.7 million unique users out of a total workforce of three million, as the department seeks to improve operational efficiency without compromising strict security requirements.

This expansion cements a new pattern in which major government institutions reshape the terms for consuming generative models through isolated intermediary platforms that assert their sovereignty over data flows.

The ChatGPT Meil model, derived from OpenAI’s Government program, focuses on delivering an experience that mirrors familiar user interfaces, including chat, file and project handling, and custom models, and is directed at processing unclassified routine work with high document density such as administrative tasks, planning, policy and supply-chain activities. In contrast, Grok for Government is offered through SpaceX’s Starshield AI arm, which relies on a secured satellite network, to cover broader operational scopes that include knowledge continuity and market-research analytics for procurement specialists and supply-chain managers in logistics.

This expansion also highlights the continued absence of Anthropic’s Claude model, following the ongoing dispute after the U.S. administration classified the company as a supply-chain risk, a decision the company is challenging in court after it refused to grant unrestricted use of its tools and insisted on specific safety controls. Meanwhile, the department has strengthened its technical partnerships with other firms, including Amazon Web Services, Microsoft, Nvidia and Reflection AI.

This shift gives digital-transformation and cyber-security officials in the governmental and sovereign sectors across the Gulf, Egypt and the region a clear operational benchmark for building shared institutional portals.

What this experiment reveals is that leveraging advanced commercial models does not necessarily require building alternative models from scratch, but instead calls for constructing unified, isolated intermediary layers that prevent model training on operational inputs while providing multiple interfaces under a single security umbrella. Government technology and procurement teams locally need to revisit commercial AI usage agreements and shift toward requiring data isolation and sovereign compliance as a core condition for adopting productivity agents in sensitive institutions.