The Pentagon's responsible AI office: 3 of 10 posts filled, a council that went dark for nine months, and still no charter

Get the provenance straight. The primary material is a U.S. Government Accountability Office report, GAO-27-107837, Artificial Intelligence: Actions and Opportunities to Implement Responsible Use in DOD, available on GAO's product page and dated October 14, 2026. Media followed from October 20: Defense News and Military Times ran pieces on October 21, and NPR published a fuller account on October 23 and 24. The report's own origin is a provision in Senate Report 118-188 accompanying the National Defense Authorization Act for Fiscal Year 2025, asking GAO to review implementation and oversight of DOD's Responsible AI guidance.

What the review covered. GAO collected and examined DOD documentation on Responsible AI, surveyed ten DOD components, selected ten specific use cases to see how six Responsible AI tenets were implemented, and interviewed cognizant officials. One finding is positive and should not be dropped: most of the surveyed components reported taking at least initial steps to establish oversight and implement the tenets. Leaving that out turns the report into a one-sided critique it is not.

Hard fact one: the council went dark. The Office of the Chief Digital and AI Officer runs a Responsible AI Division, and that division established the Responsible AI Working Council as the department's working-level collaborative group for Responsible AI. CDAO officials told GAO the council did not meet from late 2024 until August 2025. The mechanism that had served as the standing working-level platform since May 2021 was effectively empty for about nine months.

The cost of the gap is documented. Officials from four DOD components told GAO they faced Responsible AI implementation challenges during that period that collaboration would have helped resolve. Specifically, the Joint Staff described coordination problems in completing its lines of effort; the Defense Threat Reduction Agency said it wanted help interpreting federal guidance consistently for high-impact uses; the Air Force and the Navy raised their own difficulties. Separately, six of the ten surveyed components said that sharing Responsible AI resources and best practices would benefit their implementation efforts.

Hard fact two: it resumed without a charter. Officials said the council restarted meetings in August 2025, but it has no charter defining roles and responsibilities and specifying meeting frequency. GAO's judgement is that this gap leaves the council at risk of lapsing again. The consequence is stated just as concretely: another lapse could leave DOD facing implementation problems such as inconsistent interpretation of the same guidance, with no collaborative capacity to resolve them quickly.

Hard fact three: staffing. CDAO officials said that as of February 2026 only three of the ten authorized positions for the Responsible AI Division were filled. The report ties that gap directly to progress: CDAO had already identified additional actions that would help DOD implement Responsible AI, but staffing shortages hindered its ability to take them. GAO's logic is that filling the division's workforce needs is what lets CDAO make further progress toward its goal of accelerating AI across the department.

What GAO recommended. Two new recommendations, both to the Secretary of Defense. First, ensure CDAO develops and issues a charter for the Responsible AI Working Council that outlines roles and responsibilities, includes the CDAO Responsible AI Division or an equivalent office plus DOD components, and specifies how often the council meets. Second, ensure CDAO addresses the Responsible AI Division's workforce needs. DOD provided comments by email indicating concurrence with both.

One more is addressed to Congress. GAO recommends that Congress consider directing DOD to implement its open recommendations on collaboration and the AI workforce, specifically to define the roles and responsibilities of organisations and components collaborating on AI activities, and to address AI workforce challenges department-wide by assigning responsibility and setting a timeline for fully defining and identifying DOD's AI workforce. The earlier report behind this is GAO-24-105645, which had already found that DOD had not defined responsibilities for AI collaboration or addressed workforce needs affecting AI adoption.

What happened next. NPR adds a date: although DOD concurred with both recommendations, as of October 22, 2026 the changes had not been implemented, a point a GAO representative confirmed to NPR. Another widely quoted figure is that 29 of the 64 actions in DOD's Responsible AI Strategy and Implementation Pathway remained incomplete as of April 2026. That figure comes from Defense News's reading of the report rather than GAO's product summary page, so it should be attributed accordingly instead of being presented as summary text.

Three boundaries. First, this is a document review plus a survey and interviews, not a technical audit of any weapons system; the report is about governance machinery idling, not about an incident involving a specific system, and the two must not be merged. Second, the report does name use cases tied to weapon-related programmes, but those cases serve as a sample for observing how the tenets get implemented, and the report draws no compliance verdict on them. Third, concurrence is not execution, and NPR verified that as of October 22 nothing had been implemented. This article does not forecast what comes next; it records what is publicly checkable.