The Pentagon has put two of the biggest names in commercial AI on its official platform for serving its 3 million personnel. Defense One reported on Aug 31 that the Department of Defense launched OpenAI's ChatGPT Mil and xAI's Grok for Government on GenAI.mil, the department's enterprise AI platform, joining Google's Gemini for Government, which has been on the platform since its December 2025 launch. The move gives service members and civilians access to OpenAI and xAI models through a single accredited gateway, and it leaves Anthropic's Claude on the outside, still treated as a supply-chain risk after a federal judge blocked the Pentagon's earlier designation of the company.
The development is a milestone in the military's adoption of commercial AI, and it reflects a policy landscape that has been in turmoil for months. The Pentagon has been building GenAI.mil as the secure doorway through which its workforce uses large language models, and the addition of ChatGPT Mil and Grok for Government expands the platform from one major vendor to three. The absence of Anthropic is the most politically charged detail, because the company's exclusion was the subject of a court ruling, and because the reason for that ruling, that the Pentagon's security assessment was found to be illegal and baseless, remains unresolved.
Key Facts
Defense One reported on Aug 31 that ChatGPT Mil, the version of OpenAI's assistant built for military use, has been accredited at Impact Level 5, the classification that permits it to handle Controlled Unclassified Information, the most sensitive category of non-classified data. The same accreditation applies to Grok for Government, and both models are now available through GenAI.mil, which serves more than 3 million civilian and military personnel across the department. Military Times reported on Aug 31 that the platform already has more than 1.7 million active users, a figure that reflects the rapid adoption of AI tools across the force.
Quartz reported on Sep 1 that the addition of ChatGPT Mil and Grok for Government means GenAI.mil now offers models from three major AI vendors: OpenAI, xAI and Google, which has been on the platform since the platform's December 2025 launch. The department described the additions as a broadening of the platform's capabilities, giving personnel a choice of models for different tasks, from drafting and summarization to analysis and code. The Department of War said on Aug 31 that the new models are available through the same secure gateway, with the same data-handling protections, as the existing offering, and that the Impact Level 5 accreditation covers both new models for handling Controlled Unclassified Information.
The context around Anthropic is the unresolved part of the story. In August, a federal judge, Judge Rita Lin, blocked the Pentagon's designation of Anthropic as a supply-chain risk, ruling on Aug 27 that the assessment was illegal and baseless, according to reports from Defense One and Military Times on Aug 31. The judge's ruling did not order Anthropic onto the platform, but it removed the legal basis for its exclusion, and the department's decision to launch two new vendors while Anthropic remains absent means the issue is likely to return to court or to policy review. The department has not announced a timeline for reconsidering Anthropic's status.
Analysis
What this really means is that the Pentagon is treating the large AI providers as infrastructure rather than as a single trusted vendor, and the expansion of GenAI.mil from one model to three is the operational expression of that stance. A platform that offers OpenAI, xAI and Google models gives the department redundancy, competition and flexibility: no single vendor becomes indispensable, personnel can pick the best model for a task, and the department can negotiate pricing and terms from a position of strength. The accreditation of ChatGPT Mil at Impact Level 5 is the technical core of the arrangement, because it certifies that the model can handle the most sensitive unclassified information, which is the threshold that makes the platform useful for real military work.
The bigger picture here is about the tension between the military's appetite for AI and the legal and political process that governs procurement. The Pentagon clearly wants access to the best AI models, and it wants them fast, which is why GenAI.mil has grown to 1.7 million users and why the platform now spans three vendors. But the Anthropic episode shows that the process for deciding who gets to serve the military is not purely technical: it involves security assessments, court rulings and supply-chain designations, and when those collide, the result can be a vendor excluded on grounds a judge later finds unlawful. The department's decision to add two new vendors while the Anthropic question is unresolved is a reminder that the commercial AI market and the national security bureaucracy are still learning to coexist.
The expansion is good for the military and for the AI market in ways that are easy to miss. For the military, three vendors on the platform means the force is no longer dependent on a single company's models, which is a resilience gain in a domain where availability can be a matter of operational readiness. For the AI companies, the Pentagon's business is both lucrative and a powerful validation: an Impact Level 5 accreditation is a signal to every other government customer and to the private sector that a model meets the highest security standards. And for the broader market, the fact that the Pentagon is willing to run a multi-vendor platform is a sign that the government is approaching AI the way it approaches other defense suppliers, with redundancy and competition, rather than treating any single model as a permanent fixture.
Why It Matters
For the military, the launch means that most of the force now has access to three of the most capable AI models through a single accredited platform, which changes how service members draft documents, analyze intelligence and write code, and which standardizes the AI tooling available across the department. For the AI industry, the Impact Level 5 accreditation of ChatGPT Mil and Grok for Government is a landmark: it signals that the most demanding security requirements can be met by commercial models, and it opens the door to a government market that has been slow to adopt them. For Anthropic, the launch is a reminder that its exclusion remains unresolved, and that the court ruling against the Pentagon's designation has not yet translated into a place on the platform. And for the policy debate around AI and national security, the episode is a live case study in how the government balances the need for access to cutting-edge AI against the process, legal and bureaucratic, that governs who gets to provide it.
Next Up
In the coming weeks, watch for the Pentagon's response to the Anthropic question, whether it reconsiders the company's status, appeals the court ruling or lets the issue simmer, and for any announcements about additional models joining GenAI.mil. The longer-term questions are how the platform's usage evolves as the force adopts three vendors' models, and whether the multi-vendor approach becomes the template for how other government agencies procure AI. For anyone tracking the intersection of AI and national security, the near-term takeaway is that the military now runs a three-vendor AI platform, that the accreditation bar for serving the government has been reset by ChatGPT Mil and Grok for Government, and that the one major exclusion still standing, Anthropic, is a story that is far from over.
