Pentagon AI Oversight Office Lags Behind Department's Push, GAO Finds
Three of 10 authorized positions were filled as of February 2026, and 29 of 64 responsible AI efforts were incomplete as of April
By AXY Defense Desk ·
Key points
- Only three of the 10 authorized positions for the Pentagon's responsible AI mission were filled as of February 2026, according to a GAO report published Oct. 14.
- 29 of 64 efforts in the Defense Department's Responsible AI Strategy and Implementation Pathway were incomplete as of April 2026.
- The Responsible AI Working Council did not meet from late 2024 until August 2025 and still lacks a formal charter, which the GAO says leaves it at risk of lapsing again.
The Pentagon office charged with overseeing how the military uses artificial intelligence responsibly has been running well short of the staff it was authorized to have. A new report from the Government Accountability Office, the independent watchdog that audits federal programs for Congress, found that only three of the 10 authorized positions for the mission had been filled as of February 2026. The GAO report was published Oct. 14, and it lands at a time when the department is pushing to move AI into its operations faster.
The finding matters because the gap sits at the center of a tension that has been building across the Defense Department. Leaders want AI tools in the hands of troops quickly. The people tasked with checking that those tools are safe, tested and understood are stretched thin, and the watchdog says the shortfall is already slowing work that is supposed to happen.
What the watchdog found
The office at the heart of the review is the Chief Digital and Artificial Intelligence Officer, known as CDAO. The Pentagon tapped the office in 2022 to oversee responsible use of AI. Inside it sits the Responsible AI Division, which handles the day-to-day work of putting the department's principles into practice.
The GAO found that 29 of 64 efforts outlined in the Defense Department's Responsible AI Strategy and Implementation Pathway remained incomplete as of April 2026. That strategy is the department's plan for turning its stated commitments on responsible AI into specific actions. Nearly half of the listed efforts were still unfinished months into the period the watchdog reviewed.
Staffing shortages, the GAO said, had hindered the CDAO's ability to take additional steps to implement responsible AI. The department uses responsible AI, in its own description, to help accelerate AI adoption. In other words, the rules and checks are meant to be the thing that lets the department move fast with confidence, and the office that provides them is short on people.
A working council that stopped meeting
The Responsible AI Working Council, a group meant to coordinate across the department, did not meet from late 2024 until August 2025. That is a gap of many months in which the department's main forum for shared AI rules sat idle.
CDAO officials told the GAO that other priorities had kept the Responsible AI Division from organizing meetings. Those priorities included developing guidance for generative AI, the kind of technology that produces new text, images and code in response to a user's request, and contributing to government-wide AI risk-management resources. Both are substantial jobs. Generative AI in particular has raised new questions about accuracy, security and misuse, so guidance on it is not a small task.
The council resumed meeting in August 2025. However, the GAO found it still lacked a formal charter, meaning a written document defining its responsibilities and specifying how often it should meet. Without that charter, nothing in the record requires the council to keep meeting, which is why the watchdog said it remains exposed to another lapse. The GAO report reads: “Without a charter that defines roles and responsibilities for the council ... and that specifies the frequency of council meetings, the council is at risk of lapsing again.”
The watchdog also warned about the consequences of another gap. “Such a lapse could increase the risk of DOD facing Responsible AI implementation issues that it would not have the collaborative capacity to swiftly and effectively resolve,” the report reads.
The cost of the hiatus was not theoretical, according to the GAO. Officials from four DoD components, which are the department's major subordinate organizations, told the watchdog they faced implementation challenges during the pause that could have benefited from greater collaboration. Six of the 10 DoD components the GAO surveyed said that sharing responsible AI resources and best practices would increase their ability to fold the department's six Responsible AI tenets into their AI use cases.
Why staffing has become a sticking point
CDAO officials told the GAO they were unlikely to fill vacant positions or increase staffing while the office underwent restructuring. The office had been realigned to the Office of the Under Secretary of Defense for Research and Engineering, a move that changed where it sits in the department's chain of command.
The GAO pushed back on that reasoning. In its view, “by ... taking action to ensure it has staff in place to meet those needs,” CDAO could make better progress toward its goal of accelerating AI across the department. The watchdog's argument is that waiting for reorganization to finish is itself a choice with costs, because the work does not pause while the paperwork is being sorted out.
That is a familiar problem in large organizations. A restructuring can be a legitimate reason to hold off on hiring, since no one wants to fill jobs that might be moved or eliminated. But it can also leave gaps in exactly the functions that the reorganization was supposed to strengthen. The GAO's recommendation suggests it sees the second risk as the bigger one right now.
What experts say about human judgment
Owen Daniels, associate director of analysis at Georgetown University's Center for Security and Emerging Technology, a research group that studies AI and national security, said the staffing gap could make it harder to build AI literacy across the force. He argued that responsible AI and effective AI are not competing goals. “Responsible AI should be considered the same as operationally effective AI,” Daniels said.
He explained what that looks like in practice. Testing, evaluation, verification and validation are the methods used to check that a system does what it is supposed to do, under the conditions it will face. Verification asks whether the system was built correctly. Validation asks whether it is the right system for the job. Daniels said these practices help the military manage risk while making sure systems operate as intended and that operators understand their strengths and limitations.
Daniels also pointed to the people who will use the tools. “Operationally, the force needs to become AI literate, as warfighters will be responsible for determining when to deploy AI-enabled systems,” he said. Service members, he added, need to understand the limits of AI systems and recognize when handing a decision to a machine may be inappropriate, rather than defaulting to the machine's judgment.
The Pentagon uses the phrase “appropriate levels of human judgment” to describe how much control people should keep over military AI. Daniels said that idea depends on knowing the system well. “Exercising meaningful human control — or to use the department's term, appropriate levels of human judgment — over a military AI system requires robustly understanding a system's operating boundaries through development, rigorous testing and evaluation, and training,” he said.
Lawmakers raise accountability concerns
The report also touches a question that has drawn attention on Capitol Hill: who is actually in charge of AI decisions at the Pentagon. Rep. Don Bacon, R-Neb., raised a related concern about accountability during a May 2025 hearing on the department's AI posture. “It's my observation that the Defense Department doesn't lack in the number of principals, but rather, it often struggles with too little clarity in who is responsible for what.”
His point gets to the heart of the GAO's findings. Many people may have a role in AI oversight, but if responsibilities are not written down and assigned, a council, a division or a working group can stall without anyone clearly owning the problem.
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. The senators wrote that “Recent events have given us growing concern about the extent to which agencies under your oversight have prioritized acceleration of AI capability adoption and ‘experimentation’ over effective governance.” Reed has also called for Congress to set clearer guardrails for military AI.
What the GAO wants done next
The Defense Department agreed with the GAO's two recommendations. The first is to issue a charter for the Responsible AI Working Council. The second is to address the staffing needs of the Responsible AI Division, according to the watchdog.
The GAO also urged Congress to order the department to carry out earlier recommendations. Those include defining the responsibilities of organizations involved in AI activities, and assigning responsibility and a timeline for identifying and managing the department's AI workforce.
Several things are worth watching. The first is whether the department issues the council charter and sets a meeting schedule, since that would remove the risk of another lapse. The second is whether the Responsible AI Division gets the people it needs, and how quickly. The third is whether Congress acts on the GAO's request for direct orders, which would turn advice into requirements. If the department's push for faster AI adoption continues while these gaps remain open, the lawmakers who wrote to the Pentagon are likely to keep asking the same question: who is responsible for what, and are they equipped to do it.
Questions and answers
How many positions were filled in the Pentagon's responsible AI office?
As of February 2026, three of the 10 authorized positions for the Pentagon's responsible AI mission had been filled, according to the Government Accountability Office. The office is the Chief Digital and Artificial Intelligence Officer, which was tapped in 2022 to oversee responsible use of AI.
Why did the GAO say the Pentagon's responsible AI council is at risk?
The council did not meet from late 2024 until August 2025, and it still lacks a formal charter defining its responsibilities and meeting frequency. The GAO warned that without a charter, the council is at risk of lapsing again.
What did the Defense Department say about the GAO recommendations?
The Defense Department agreed with the GAO's two recommendations, which are to issue a charter for the Responsible AI Working Council and to address the workforce needs of the Responsible AI Division. The GAO also urged Congress to order the department to implement earlier recommendations.
Why does staffing matter for military AI?
Georgetown researcher Owen Daniels said the shortfall could make it harder to build AI literacy across the force, since service members need to understand AI systems' limits. Effective testing and evaluation are part of managing risk, and the GAO said staffing is needed to meet those needs.
Topics: Pentagon, Government Accountability Office, Responsible AI, Chief Digital and Artificial Intelligence Office, Defense Department, Military artificial intelligence, Congress AI oversight
Sources: Military Times