James Mazol, deputy undersecretary of defense for research and engineering, said Defense Department’s (DoD) AI adoption is accelerating, according to a DefenseScoop report. 

Sharing the update on the September 22 DefenseTalks conference, he said adoption grew this fast largely because the department consolidated personnel onto one shared platform, GenAI.mil. 

About 500,000 of its 1.7 million users were now “power users,” relying on it daily to do their jobs, said Cameron Stanley, the Pentagon’s chief digital and AI officer (CDAO).

According to DefenseScoop, the growth stands in sharp contrast to where the department started. 

Mazol said that when the second Trump administration began, only about 80,000 people across the department were using generative AI, relying on homebrew models built at the Air Force Research Laboratory (AFRL) and similar sites.

More Users, More Classified Networks, Less Capacity

GenAI.mil was launched in December 2025 by DoD (also known as Department of War) to bring advanced generative AI tools to the entire military workforce. 

The Army, Navy, Marines, Air Force, and Space Force have all since adopted it as their preferred enterprise AI system. Grok and ChatGPT joined Gemini on the platform last month.

Mazol said the department now plans to extend frontier models, the most advanced AI systems currently available, beyond GenAI.mil’s current controlled unclassified network. 

According to the report, the department’s plan includes: 

• SIPR (the Secret Internet Protocol Router Network, used for secret-level classified information)
• JWICS (the Joint Worldwide Intelligence Communications System, used for top-secret information) 
• Special Access Programs, which restrict access to the department’s most sensitive projects

Stanley told the same conference the department was now working out how to integrate AI agents and map agentic workflows, automated processes where AI systems act on tasks with limited human input, according to the report. 

But he also flagged a real constraint ahead. “AI is getting put into everything. That scares me, mainly because of capacity. We don’t have the capacity,” he said. 

Scale Without a Matching Plan Creates Its Own Risk

For chief data officers (CDOs), the Pentagon’s experience is a preview of a problem that hits any organization once AI adoption moves fast enough: growth outpaces the infrastructure meant to support it. 

Going from 80,000 to 1.7 million users, and now pushing frontier models into classified environments, is quite an achievement, but Stanley’s own warning about compute capacity is something to focus on here.

It’s the same discipline behind broader AI governance work: scaling AI use responsibly means matching capability to actual need, instead of just expanding access as fast as demand grows.