"It's just a really exciting time for generative AI in the department," Cameron Stanley, the chief digital and artificial intelligence officer at the Department of Defense, told attendees at the AWS Summit in Washington, D.C.
GenAI.mil’s scale: 1.7 million users and 100,000 custom agents
GenAI.mil, the Defense Department’s internal artificial intelligence marketplace, has reached a reported record of nearly 1.7 million users and the creation of more than 100,000 custom agents, the department announced at the AWS Summit. That scale reflects widespread internal uptake of tools and workflows hosted on a single, department-wide platform.
The platform already hosts capabilities from a range of commercial vendors: SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Oracle, and Amazon Web Services. Those offerings are available at Impact Level (IL) 6 and 7, a classification the Pentagon publicized in May.
New model additions and plans to reach higher classification levels
Stanley said the department is planning to put additional models onto GenAI.mil and to extend the platform to “higher classification levels.” He framed those moves as part of a broader procurement and deployment strategy: “We're looking forward to advancing, getting new models on to GenAI.mil, we're looking at GenAI.mil going to higher classification levels,” he said.
One concrete near-term change already disclosed: OpenAI confirmed in mid‑June that its flagship chatbot, ChatGPT, will be eligible for controlled, unclassified information through GenAI.mil beginning in July.
Commercial‑first procurement and the vendor‑near‑warfighter model
Stanley described the overarching policy direction as “commercial‑first.” He said the aim is to place vendors alongside operational users so that a single, clear objective—delivering what the warfighter needs—drives development and delivery.
“We're trying to put the vendor next to the warfighter and have the vendor have one goal, one job, that's it, and that is to deliver exactly what the warfighter’s needs are,” Stanley said. “What we do is we create the environment with the right tools and the right environments with the right security in place with the right contracts in place.”
Agentic tools, tight guardrails, and faster analysis for decision-making
Stanley emphasized that the department is deliberately adding “agentic” tools to support analytics, accompanied by “very tight guardrails.” He argued the technology addresses a core operational problem: humans cannot always parse the volume of data required on modern battlefields at necessary speed.
Describing the efficiency gains, he said the platform’s agentic analytics replace workflows that previously required multiple systems and several analysts. Stanley estimated the output equates to the work of “two to three human analysts operating in disparate systems,” and described a transition “instead of having six or seven systems we have to go across in order to make that decision—we're now doing it instantaneously, or nearly instantaneously, with humans appropriately managing the entire workload process and actioning it from the same system that we identified the decision from.”
What this means for technologists, procurement leaders, and warfighters
- Technologists and security teams: Expect expansion of vendor integrations already running at IL‑6 and IL‑7, and preparation for higher classification deployments; engineering and security work will need to align with the “very tight guardrails” Stanley described.
- Procurement leaders and policy teams: The “commercial‑first” posture signals a continued emphasis on contracting and environment design—security, tools, and contracts must be in place to let vendors operate near end users.
- Warfighters and operational commanders: The department is prioritizing agentic analytics that aggregate data and accelerate identification of critical information, with humans remaining responsible for managing and actioning decisions from within a consolidated system.
GenAI.mil’s reported user numbers, vendor roster, upcoming ChatGPT eligibility for controlled unclassified data in July, and explicit move toward higher classification levels together outline a concrete, department-level push to normalize commercial generative AI across a broad set of missions. The immediate next steps Stanley signaled are clear: more models onto the marketplace and expansion to higher classification domains—alongside the contractual and security work to support them. How quickly and which models will follow ChatGPT into broader classified work are the proximate operational questions the department has set in motion.
