The Pentagon office tapped in 2022 to oversee the navy’s accountable use of artificial intelligence had crammed merely three of its 10 licensed positions for the mission as of February 2026, even as a result of the division pursues a way to pace up AI development, a authorities watchdog found.
The Authorities Accountability Office discovered that 29 of 64 efforts outlined inside the Safety Division’s Accountable AI Approach and Implementation Pathway remained incomplete as of April 2026.
In accordance with a report revealed Oct. 14, the GAO moreover found that staffing shortages had hindered the Chief Digital and Artificial Intelligence Officer’s functionality to take further steps to implement accountable AI, which the division makes use of to help pace up AI adoption.
GAO advisable that the division cope with staffing needs in its Accountable AI Division, which falls beneath the Chief Digital and Artificial Intelligence Office, and concern a correct structure for the working council.
“Accountable AI must be thought-about the similar as operationally environment friendly AI,” talked about Owen Daniels, affiliate director of research at Georgetown School’s Coronary heart for Security and Rising Know-how.
Environment friendly testing, evaluation, verification and validation practices would possibly assist the navy deal with risks whereas guaranteeing packages operate as supposed and operators understand their strengths and limitations, Daniels talked about.
Council ‘liable to lapsing’
The GAO found that the Accountable AI Working Council did not meet from late 2024 until August 2025.
Chief Digital and Artificial Intelligence Officer, or CDAO, officers instructed the watchdog firm that completely different priorities, along with creating steering for generative AI and contributing to government-wide AI risk-management sources, had prevented the Accountable AI Division from organizing conferences.
Rep. Don Bacon, R-Neb., raised a related concern about accountability all through a May 2025 listening to regarding the Pentagon’s AI posture.
“It’s my assertion that the Safety Division doesn’t lack inside the number of principals, nonetheless fairly, it often struggles with too little readability in who’s liable for what.”
Officers from 4 DoD components instructed the GAO they confronted implementation challenges by the hiatus that might have benefited from bigger collaboration.
Six of the ten DoD components surveyed talked about sharing accountable AI sources and best practices would enhance their functionality to incorporate the division’s six Accountable AI tenets into their AI use cases.
The council resumed conferences in August 2025, nonetheless the GAO found it nonetheless lacked a correct structure defining its duties and specifying how incessantly it ought to fulfill.
“And never utilizing a structure that defines roles and duties for the council … and that specifies the frequency of council conferences, the council is liable to lapsing as soon as extra,” the GAO report reads. “Such a lapse would possibly enhance the possibility of DOD going by Accountable AI implementation factors that it couldn’t have the collaborative functionality to swiftly and efficiently resolve.”
Sluggish staffing
CDAO officers instructed the GAO they’d been unlikely to fill vacant positions or enhance staffing whereas the office underwent restructuring, following its realignment to the Office of the Beneath Secretary of Safety for Evaluation and Engineering.
In response, the GAO argued that “by … taking movement to ensure it has workers in place to satisfy these needs,” CDAO would possibly make larger progress in direction of its objective of accelerating AI all through the division.
Daniels talked about the staffing shortfall would possibly make it harder to assemble AI literacy all by the drive.
“Operationally, the drive should turn into AI literate, as warfighters will most likely be liable for determining when to deploy AI-enabled packages,” he talked about.
Service members wish to know AI packages’ limitations and acknowledge when delegating selections to machines is also inappropriate, fairly than defaulting to a machine’s judgment, Daniels talked about.
“Exercising important human administration — or to utilize the division’s time interval, acceptable ranges of human judgment — over a navy AI system requires robustly understanding a system’s working boundaries by development, rigorous testing and evaluation, and training,” he talked about.
Congress’ place
The DoD agreed with the GAO’s two recommendations to concern a structure for the Accountable AI Working Council and cope with the Accountable AI Division’s workforce needs, in accordance with the watchdog.
The GAO moreover urged Congress to order the division to implement earlier recommendations, along with defining the duties of organizations involved in AI actions and assigning responsibility and a timeline for determining and managing the division’s AI workforce.
In a Sept. 21 letter, Sens. Jack Reed, D-R.I., Mark Warner, D-Va., and Chris Coons, D-Del., expressed concern that the Pentagon’s AI push could be outpacing its governance efforts.
“Newest events have given us rising concern regarding the extent to which companies beneath your oversight have prioritized acceleration of AI performance adoption and ‘experimentation’ over environment friendly governance,” the letter reads.
Reed has moreover known as for Congress to find out clearer guardrails for navy AI.
“Congress ought to do its job and provide balanced oversight, necessities, and tips of the freeway,” he talked about.
Natalie Oliverio is an unbiased journalist and Navy veteran who covers navy and veterans’ factors. Drawing on her experience as a experience authorities and mentor, she writes regarding the insurance coverage insurance policies, people and functions shaping the navy neighborhood.
