The Pentagon has dramatically expanded its AI arsenal. On August 31, 2026, the Department of Defense announced that two frontier AI models — OpenAI's ChatGPT Mil and Starshield AI's Grok for Government — are now operational on GenAI.mil, the military's secure enterprise portal for generative AI.
The additions join Google's Gemini for Government, which launched on the platform in December 2025 and has since attracted over 1.7 million users. Together, the three models give more than 3 million military members, civilian employees, and contractors access to a diverse suite of AI capabilities for unclassified work.
ChatGPT Mil will support "document-heavy unclassified work across the Department, including planning, policy, logistics, and administration," the Pentagon said. A source familiar with the integration told DefenseScoop that users can currently interact with GPT-5.4 Terra, with GPT-5.6 Terra — OpenAI's latest frontier model — "sequenced" for release soon.
Grok for Government, meanwhile, offers "deep-thinking inference, adaptive reasoning modes (Auto, Fast, and Expert), customizable workspaces, persistent projects and reusable 'playbooks' that capture and scale institutional knowledge," per the Pentagon's announcement.
All three models have cleared Impact Level 5 (IL5), the highest authorization tier for handling sensitive unclassified data. The Pentagon framed the expansion as part of its "wartime speed" approach to AI adoption.
The rollout carries political and legal baggage. In February, the Pentagon designated Anthropic — maker of Claude — a national security supply-chain risk after the company insisted on contractual guardrails preventing use of its AI in lethal autonomous weapons or mass surveillance of Americans. Anthropic sued, and on August 22, a federal district judge ruled the Pentagon's actions were "illegal and baseless."
Pentagon CTO Emil Michael said last week that the department expects to complete its full removal of Anthropic products from national security systems by the end of September. The absence of Claude on GenAI.mil leaves a visible gap in the platform's model roster.
The integration also comes as Grok faces ongoing scrutiny over safety and reliability issues in its consumer-facing product. Despite this, the Pentagon has doubled down on its value: Cameron Stanley, the DoD's chief digital and AI officer, stated in a court filing that Grok enabled U.S. forces to deploy over 2,000 munitions to 2,000 targets within 96 hours during the Iran conflict.
For now, the Pentagon's message is clear: it wants multiple AI providers to avoid vendor lock-in, and it's willing to fight — in court and in Congress — for that flexibility.
