Sunday, September 6, 2026

News

Pentagon Adds ChatGPT and Grok Alongside Gemini for Millions of Employees

MarketPatryk Raba
Pentagon Adds ChatGPT and Grok Alongside Gemini for Millions of Employees
Fot. Chief Photographer's Mate Johnny Bivera, U.S. Navy, Wikimedia Commons (Public domain)

The US Department of War added military versions of ChatGPT and Grok to its GenAI.mil platform, joining Gemini. The tools are now available to more than 3 million soldiers and civilian employees, with 1.7 million already logged in.

Contents
  1. What was added
  2. Data stays in the government cloud
  3. Anthropic still absent
  4. What this means for the market

The US Department of War expanded its internal artificial intelligence platform GenAI.mil on Monday to include military versions of ChatGPT and Grok. Until now, Google's Gemini was the only model available there; now soldiers and civilian employees of the department can choose between three competing AI systems within a single environment isolated from the internet.

What was added

ChatGPT Mil is a military version of OpenAI's product, including chat, file handling, projects, and custom model configurations, designed mainly for office work: planning, policy drafting, logistics, and administration. The company also says it will soon add the newer GPT-5.6 Terra model in the same version. Grok for Government, from Starshield AI, xAI's subsidiary serving the government sector, offers deeper reasoning modes, configurable workspaces, and reusable task templates.

Both models join Gemini, which was the first to arrive on the platform in December 2025 and quickly became the default AI tool in five of the six branches of the US armed forces. All three passed accreditation at Impact Level 5, meaning they are cleared to work with unclassified but sensitive Controlled Unclassified Information (CUI), though not with fully classified data.

Data stays in the government cloud

The department's key argument is that GenAI.mil is kept separate from the commercial versions of the same chatbots. OpenAI has stated that data processed within GenAI.mil remains isolated within the government environment and is not used to train commercial models. The platform also prohibits processing personal data and medical information, intended to protect against leaks of sensitive data belonging to soldiers and employees.

We're moving at wartime speed: we're enabling the department to experiment, iterate, and deploy faster than adversaries can adapt. - an anonymous US Department of War official
We want to combine access to powerful tools with clear safeguards and sound human judgment. - Joe Larson, Vice President and Head of Government at OpenAI

Anthropic still absent

Anthropic, the maker of Claude, is still missing from the list of providers, despite having sought a place on US government platforms for months. In February 2026, the Pentagon designated Anthropic a supply chain risk and removed the company's products from its systems. Anthropic challenged the decision, and a week before GenAI.mil's expansion to include ChatGPT and Grok, a federal judge ruled that the Pentagon's actions against the company had been illegal and unfounded. It remains unclear whether the ruling will open the door for Claude to join the platform.

What this means for the market

For OpenAI and xAI, joining GenAI.mil is a major win both reputationally and financially: access to a base of more than 3 million potential users within one of the world's largest organizations, alongside Google's existing Pentagon contract. It's also a signal that competition among large language model providers is shifting increasingly toward the government and defense procurement market, where security accreditations matter as much as, if not more than, benchmark performance.

For Polish companies and institutions, the GenAI.mil story points to a direction public administration may follow: instead of choosing a single AI vendor, more organizations are opting for multi-model platforms that let employees test several systems in parallel within a controlled, isolated environment. In Poland, oversight of AI in public administration and the military is still in its early stages, with the national artificial intelligence commission set to begin work in November.

The scale of adoption within the US military, 1.7 million users in less than a year since launch, also shows how quickly large organizations can integrate generative AI into everyday office work once data security concerns are addressed. The example of the Airborne Corps, which cut the preparation of an operations order from several months to six weeks, will likely be cited as an argument for similar rollouts in other NATO armies.

Share: