WASHINGTON — Deputy Secretary of Defense Steve Feinberg has ordered a major overhaul of the military’s artificial intelligence efforts, including a reorganization that some former officials worry could be seen as a demotion for the Pentagon’s main AI office.
Feinberg’s Aug. 14 memo launches three major reviews over the next 60 to 120 days and subordinates the Chief Digital & AI Office (CDAO) — which had reported directly to Feinberg — to the Pentagon’s new R&D chief, Emil Michael, “effective immediately.”
“This realignment will not create additional review layers or bureaucratic processes,” a defense official told Breaking Defense after confirming the change. “CDAO will continue to execute all current statutory responsibilities without interruption during this transition.”
However, it’s not specified how long “this transition” will last or what comes next. It may be tied to the timeline for the reviews, as the memo gives Michael 120 days to deliver “a recommended path forward” for two high-priority AI-driven data-sharing systems currently run by CDAO and widely used at high levels in DoD, Advana and the Maven Smart System.
The memo, first reported by Defense Scoop, also gives Michael three months to recommend “updates” to DoD Directive 5015.98 [PDF], the founding charter that lays out all of CDAO’s authorities and responsibilities. The R&D chief, whose formal title is under secretary of defense for research and engineering (USD(R&E)), has just 60 days to deliver “a comprehensive DoD AI strategy” aligned with President Trump’s AI Action Plan.
CDAO was created in late 2021 from a merger of multiple high-tech initiatives to coordinate AI and big-data improvements across the Department of Defense. It has authorities for policy, R&D, and procurement — three functions which are normally divided among three different under secretaries. It reported directly to the deputy secretary, who traditionally handles day-to-day management of the Department.
“This realignment is the next step in making a uniform, AI-first push for the Department of Defense,” the defense official said. “By aligning the CDAO under the USD(R&E), we create a powerful innovation engine that can deliver AI superiority from laboratory to battlefield.”
One former official, who asked not to be identified, suggested the move could be an advantage if it meant a more robust institutionalization of AI.
“If this signals the mainstreaming of AI within Department processes and functions, that is great,” the ex-official told Breaking Defense. “Industry is already mainstreaming AI, [so] it is appropriate for the Department to do so as well. Every service has an AI component today, putting implementation into the hands of warfighters.”
On the other hand, the same former official continued, moving AI under R&D “seems to fly in the face” of the need for rapid deployment on a wide scale.
“The Department should be scrambling to implement the technology across workflows in every warfighting function,” they said. “Our competitiveness is in the balance, and our opponents are already using advanced autonomy on the battlefield. The U.S. does not have the leeway to delay the maturation of ‘real’ warfighting advantages.”
Other former officials also said that AI is so new to most of the military that it still requires special treatment, including direct access to top levels of leadership.
“This reorganization would make more sense if AI integration and scaling was much further along in the Department,” said retired Air Force Lt. Gen. Jack Shanahan, who led the pioneering Project Maven and one of CDAO’s predecessor organizations, the Joint AI Center.
“An administrative reorganization like this does not suggest that AI adoption across the DoD is as urgent as senior leaders have suggested,” Shanahan told Breaking Defense. “When you pull an organization that was a direct report to the deputy secretary or secretary and move it somewhere else in the Pentagon, no matter what the intent might be, the message to the force is loud and clear: This isn’t a priority.”
In particular, Shanahan said, the memo does not mention what will happen to CDAO’s current connections to the operational Combatant Commands around the world. The COCOMs have worked closely with CDAO to both field-test and deploy cutting-edge technology, including an early version of the AI-driven command-and-control system known as CJADC2.
A former DoD policy official, Michael Horowitz, agreed with Shanahan that folding CDAO under R&D seemed to be a step backward, away from the Pentagon’s ostensible goal of deploying AI at scale, ASAP, across the armed services.
“Demoting AI within the Pentagon seems pretty risky at this point in history,” Horowitz told Breaking Defense. “The success of this move will ride on the extent to which AI has achieved escape velocity within the services [already] … such that you don’t need a super-empowered AI organization. [But] the Pentagon’s already been too slow when it comes to scaling AI adoption.”