Pentagon Adds ChatGPT and Grok to Its Internal Platform

The Pentagon has added OpenAI's ChatGPT and xAI's Grok to its internal generative AI platform, giving millions of US military and civilian personnel access to multiple AI models.

ChatGPT Mil and Grok for Government have joined Google's Gemini on GenAI.mil, a centralised security platform designed to give Department of Defense personnel access to commercial AI tools.

This environment has been designed to handle classified data securely in a way that is not accessible to the public. TechCrunch reports that the additions now give Pentagon employees access to more competing AI models through a single government-sanctioned platform.

The move expands the Pentagon's use of generative AI while creating a multi-model environment in which personnel can access different AI systems through the same platform.

Pentagon Expands Access to Enterprise AI

GenAI.mil was launched to give the Department of Defense (now also known as the Department of War) personnel access to generative AI capabilities without relying on consumer-facing AI services for government work to prevent leaking sensitive information.

The platform has seen significant adoption. From more than three million military and civilian personnel, around 1.7 million unique users access it regularly, according to official government reports.

ChatGPT Mil has been specifically designed to support tasks such as planning, policy, logistics and administration, while simultaneously providing the familiar ChatGPT interface in a secured government environment.

The Department of Defense says ChatGPT Mil is accredited to handle Controlled Unclassified Information (CUI) at Impact Level 5, a security standard covering sensitive but unclassified government information.

Grok for Government has also been added to the platform, expanding the range of commercial AI capabilities available to personnel.

The additions are part of a broader effort by the Pentagon to integrate commercially available AI technology into its operations, while also maintaining appropriate security controls on how government information is accessed and used.

A Multi-Model Approach to Military AI

With ChatGPT and Grok arriving on the scene, the Pentagon is no longer putting all its eggs in one AI vendor’s basket. GenAI.mil now brings together AI models from Google, OpenAI and xAI, giving users access to multiple systems through the same platform.

This approach is relevant to all organisations that are increasingly dependent on AI, since having access to more than one model can provide greater resilience. Running into a problem with one AI provider will not result in every AI capability becoming unavailable and one can switch between models based on what tasks they’re better suited to.

The approach can also reduce dependence on a single vendor as AI becomes more deeply embedded in everyday operations. Rather than building an entire AI strategy around one model, organisations can maintain access to alternatives as capabilities, costs and performance change.

However, managing multiple AI systems introduces its own challenges. Organisations need to appreciate how different models access data, how users are authorised to interact with them, and how their outputs are monitored.

As enterprises develop their AI governance strategies, questions around vendor control, human oversight, security and risk become increasingly important.

The Pentagon's approach also demonstrates why AI is becoming part of organisations' wider technology architecture, rather than simply another standalone application. As multiple AI systems become embedded in business processes, the infrastructure, data and permissions supporting them become increasingly interconnected.

Security Remains a Concern

Increasing the availability of multiple AI platforms also raises new security issues. Providing millions of users (though they may be military) with access to AI tools requires establishing clear controls around what these information models are allowed to access and what users are permitted to do with it once they access that information.

This becomes particularly important as AI systems evolve from tools that simply generate information into systems that can interact with enterprise applications and data as humans do. For example, AI agents can now change enterprise data rather than simply read it, creating new questions around permissions, accountability and oversight.

The Expansion of the Pentagon's AI Strategy Continues

The addition of ChatGPT and Grok is part of Pentagon casting a wider net to increase its use of commercial AI technology.

The department has established partnerships with a range of technology companies as it builds out its AI capabilities as well as GenAI.mil providing a central environment through which personnel can access these approved AI tools.

The move does come amid an ongoing dispute between the Pentagon and Anthropic over the use of their Claude models for military purposes. Anthropic has sought restrictions around how its technology is used, specifically when it comes to mass surveillance and autonomous weapons, while the Pentagon wants broader access.

The dispute led the administration labelling Anthropic a supply-chain risk, a decision that has since faced legal challenges around the company’s potential blacklisting. This shows up in the fact that Claude is notably absent from the current GenAI.mil line-up, while models from their competitors Google, OpenAI and xAI are available.

For organisations adopting generative AI, the Pentagon's approach illustrates how enterprise AI is evolving beyond the question of which model is best. As more organisations adopt multiple AI systems, they will also need to manage the governance, security, data access and infrastructure surrounding them.