GAO Finds Pentagon’s Responsible AI Office Filled 3 of 10 Jobs

The Pentagon wants to put drones, robots and artificial intelligence at the center of how America fights. The small team responsible for making sure that technology is used safely and lawfully is running on less than a third of its staff.

That is the core finding of a Government Accountability Office report released Oct. 14, titled “Artificial Intelligence: Actions and Opportunities to Implement Responsible Use in DOD.” Auditors found the Pentagon’s Responsible AI Division had filled 3 of its 10 authorized positions as of February 2026.

The report surfaced the same day senior Pentagon leaders took the stage at the Association of the U.S. Army convention in Washington to sell their plan for an Autonomous Warfare Command. The timing puts two questions side by side: how fast the military can field autonomous weapons, and who is checking the work as it does.

What GAO Found Inside the Pentagon Responsible AI Effort

The Pentagon responsible AI framework has been on paper for years. The department adopted AI ethical principles in 2020 and later defined six tenets to guide how it builds, tests and fields AI systems. Its Responsible AI Strategy and Implementation Pathway, released in 2022, laid out dozens of specific tasks.

GAO says execution has lagged. As of April 2026, 29 of the 64 efforts in that pathway were still incomplete, Defense News reported in its review of the audit.

The study was requested by the Senate Armed Services Committee in the report accompanying the fiscal 2025 defense authorization bill. Auditors reviewed department documents, interviewed officials and surveyed 10 Pentagon components on how they applied the responsible AI tenets across 10 selected AI use cases.

Most components reported taking at least initial steps to set up oversight. The trouble sits higher up, in the office meant to coordinate them.

The council that stopped meeting

The Responsible AI Working Council is supposed to be the place where the services and defense agencies compare notes. According to GAO, it did not meet from late 2024 until August 2025.

Officials from four components told auditors they ran into implementation problems during that gap that better coordination could have eased. Six of the 10 components said sharing responsible AI resources and best practices would help them.

Officials in the Chief Digital and Artificial Intelligence Office, known as CDAO, said other priorities crowded out the meetings, including guidance on generative AI and governmentwide risk-management tools. The council restarted in August 2025, but it still has no charter that spells out who does what or how often it meets. GAO wrote that the group “is at risk of lapsing again.”

Three people where ten were planned

The staffing number is the one that jumps off the page. With 3 of 10 positions filled, GAO found the division could not take additional steps to put the Pentagon responsible AI tenets into practice.

CDAO officials told auditors they were unlikely to fill those vacancies while the office restructures following its move under the Under Secretary for Research and Engineering. GAO’s response was blunt in substance: the department relies on CDAO to speed AI adoption, and it cannot do that well without the people to do the job.

What GAO Recommends and How the Pentagon Responded

GAO made two new recommendations to the Secretary of Defense:

• Have CDAO write and issue a charter for the Responsible AI Working Council that defines roles, responsibilities and meeting frequency.
• Have CDAO address the Responsible AI Division’s workforce needs.

The department concurred with both. That matters, because a concurrence gives Congress and auditors a benchmark to check later.

GAO also went over the department’s head. It asked Congress to consider directing the Pentagon to carry out older, still-open recommendations, including defining which organizations own which AI responsibilities and setting a timeline for identifying the department’s AI workforce. When an audit agency asks lawmakers to step in, it usually means earlier advice has sat unaddressed for some time.

The Autonomous Warfare Command Raises the Stakes

The audit would be a routine management story if the Pentagon were standing still. It is not.

On Sept. 30, speaking to about 600 junior officers and noncommissioned officers at Marine Corps Base Quantico, Hegseth announced a new four-star Autonomous Warfare Command with “service-like authorities” to scale robotic and autonomous capabilities across the joint force, according to the Army’s account of the State of the Force address. He called the planned shift “the fastest peacetime shift in modern history.”

A memo released after the speech, titled “Project Agincourt: Pathway to the Autonomous Warfare Command,” set a target of Oct. 1, 2027, for the command to stand up, Defense One reported. It would be the military’s 12th combatant command. Stars and Stripes reported the memo says the Pentagon must work with Congress on legislation to create it.

Project Agincourt takes over the existing office for unmanned offensive and defensive systems, led by Owen West, and is charged with prototyping a faster “warfighting acquisition” model that moves decisions and money closer to front-line units. The Army, Navy and Air Force are each building new job specialties for autonomy.

The Army moved first. Acting Army Secretary Adam Telle directed a new Futures and Autonomous Systems Command the day after Hegseth’s speech, and Gen. Christopher LaNeve said Oct. 13 it will be led by a three-star general and start building autonomous systems at the corps level.

At the Oct. 14 convention panel, chief Pentagon spokesman Sean Parnell described the initiative as “not another procurement effort.” Hegseth said the work would keep iterating, adapting and learning, and moving faster.

Timeline: Pentagon Responsible AI and the Autonomy Push

• 2020: The department adopts its AI ethical principles.
• 2022: The Responsible AI Strategy and Implementation Pathway lays out 64 specific efforts.
• Late 2024 to August 2025: The Responsible AI Working Council does not meet.
• February 2026: 3 of 10 Responsible AI Division positions are filled.
• April 2026: 29 of the 64 pathway efforts remain incomplete.
• Sept. 30, 2026: Hegseth announces the Autonomous Warfare Command at Quantico.
• Oct. 14, 2026: GAO releases its report as Pentagon leaders pitch the new command at the AUSA convention.
• Oct. 1, 2027: Target date to stand up the command, pending legislation.

Two Sides of the Speed Debate

Supporters of the push make a serious case. Drones have reshaped the battlefield in Ukraine and the Middle East, and China is investing heavily in military AI. GAO itself notes that the department’s 2026 AI strategy emphasizes speed because adversaries are moving fast. From that view, a slow, paperwork-heavy oversight process is a risk of its own, and putting autonomy under one four-star command creates clear ownership.

Skeptics argue that speed without guardrails is how expensive mistakes happen. In a September letter, Sens. Jack Reed of Rhode Island, Mark Warner of Virginia and Chris Coons of Delaware, all Democrats, raised concern that AI adoption may be outpacing governance and asked inspectors general to investigate a reported AI-assisted intelligence error. According to news reports, an AI-generated assessment wrongly flagged a Chinese ship as carrying nuclear-weapons components to the Middle East.

Concern about accountability is not confined to one party. At a May 2025 House hearing on the Pentagon’s AI posture, Rep. Don Bacon, a Nebraska Republican and retired Air Force brigadier general, said the department’s problem was less the number of officials involved and more unclear accountability for who handles what, according to Defense News.

Owen Daniels of Georgetown’s Center for Security and Emerging Technology told Defense News that responsible AI and effective AI should be treated as one goal, because good testing helps operators understand what a system can and cannot do. He warned staffing gaps could slow AI literacy across the force.

Why It Matters

For the troops who will operate these systems, responsible AI is not an abstraction. It is the testing that tells a drone operator when the targeting software is reliable, and the training that keeps a young sergeant from trusting a machine’s call by default.

It also shapes accountability. When an autonomous system gets something wrong, someone has to own the result. A council with no charter and a division at 30 percent strength make it harder to say who that someone is before the mistake happens, rather than after.

For military families and veterans, the shift is a career story too. New autonomy job fields mean new training pipelines, new promotion paths and new skills that carry into civilian work. Stucci Media has covered how quickly drones are changing security at home, including the new drone ban along the U.S.-Mexico border, and the same technology now sits at the center of the Pentagon’s force design.

For taxpayers, the question is value. A new combatant command, new specialties and a new acquisition model will cost money. GAO’s work gives Congress a ready yardstick: if the department cannot staff a 10-person oversight office, lawmakers will want to know how it plans to manage a four-star command.

What Happens Next

Three things are worth watching over the coming year.

• The charter. The Pentagon agreed to write one for the Responsible AI Working Council. GAO typically tracks whether agencies follow through.
• The staffing. Whether the Responsible AI Division fills its seven empty positions after CDAO’s restructuring will be a concrete test of the department’s commitment.
• The legislation. Lawmakers will have to decide whether and how to authorize the Autonomous Warfare Command before the Oct. 1, 2027, target. That debate gives Congress a natural place to write oversight requirements into law, as GAO suggested.

As of Oct. 22, Stucci Media found no public timeline from the department for completing the remaining pathway efforts. For more coverage of force structure and defense policy, see our Military section.