Inside Cybercom’s AI Playbook: Fast Failures, Quiet Risks, and a $5M Experiment

In the massive machinery of the Pentagon’s $1.3 billion cyber R&D budget, a modest $5 million line item might seem like a rounding error. But buried in U.S. Cyber Command’s fiscal 2026 request is something that could reshape how the military fights its cyber wars: a dedicated AI program aimed at accelerating—and testing—the use of artificial intelligence for frontline cyber operations.

The initiative, known as Artificial Intelligence for Cyberspace Operations, is Cybercom’s latest move in response to a congressional directive from 2023, which pushed the command and its partners—including DARPA, the NSA, and the Chief Digital and AI Office—to rapidly develop and implement an AI roadmap for cyber warfare. The result: a roadmap, a task force, and now a pilot program embedded in one of the most operationally critical arms of U.S. cyber defense—the Cyber National Mission Force (CNMF).

“I think that the Cyber National Mission Force housing the AI Task Force makes a lot of sense,” said Gabrielle Hempel, Security Operations Strategist and Threat Intelligence Researcher at Exabeam. “The CNMF is essentially a hub for cyber operations, and they’re best positioned to understand real-world applications and identify where AI can speed up decision loops or close detection-response gaps.”

Cybercom’s AI ambitions are specific and targeted. The new program is tasked with developing data standards, tagging and curating datasets, and piloting AI-driven solutions for vulnerabilities and exploits, network security, predictive analytics, identity attribution, and infrastructure monitoring. It’s less about moonshot algorithms and more about practical, battlefield-grade AI.

Central to the plan is a 90-day agile pilot cycle, where small teams within the CNMF will test new tools against live operational scenarios—either proving them useful or moving on. That fast tempo is key, according to Hempel: “With how quickly AI is changing (and how quickly things move in cybersecurity) this is the tempo we need to meet or exceed. A long-term waterfall development would be a nonstarter.”

Still, questions loom. To fund the new initiative, Cybercom is repurposing dollars from its operations and maintenance budget rather than receiving dedicated top-line funding. That has some experts worried about long-term sustainability.

“We all know that developing, hosting, and using AI is expensive,” Hempel warned. “If Cybercom is borrowing from O&M and not getting top-line growth for this, I’m not sure what sustainability and prioritization look like across the DoD’s broader AI strategy.”

Then there’s the issue of interagency coordination. While the AI roadmap was meant to be a joint effort spanning the Defense Department’s top AI and research entities, there’s limited public insight into how those collaborations are being carried out—or whether they’re even working.

“AI innovation in the DoD can’t be siloed, and Cybercom’s approach here will either reinforce or further fracture integration across those orgs,” Hempel noted.

Finally, as Cybercom leans on commercial AI tools to bolster its capabilities, the risks multiply. Proprietary algorithms, hidden biases, and supply chain vulnerabilities all become potential weak points—especially if the systems being piloted aren’t rigorously tested.

“The Cyber Immersion Lab needs to be rigorously red-teaming and testing these tools, not just piloting them,” Hempel emphasized. “Zero trust!”

As Cybercom barrels into the AI age on a lean budget and tight timeline, the program will serve as a high-stakes experiment—not only in technology, but in how agile and aligned the Pentagon can be when deploying AI at cyber speed.