The Department of Defense is developing weapons and systems with artificial intelligence capabilities that do not consistently adhere to the Pentagon’s own guidelines for responsible AI, according to a report published by the Government Accountability Office (GAO) in October. The survey of ten defense projects found that officials frequently identified risks associated with using these tools in operational environments, yet several teams were not following the department's established ethical tenets. The findings come as Defense Secretary Pete Hegseth prioritizes rapid autonomous innovation, declaring the department an "AI-first" warfighting force.
The GAO report, released on Oct. 14, examined ten specific projects among more than 300 AI initiatives currently under development at the Pentagon. These projects include systems for counter-camouflage, target recognition, surveillance processing, and tactical decision-making. While the Department issued a memo in 2021 outlining six tenets for responsible AI—emphasizing oversight, verification, real-time monitoring, and cooperation with allies—the implementation of these guidelines appears inconsistent. Alissa Czyz, who directed the GAO review, noted that while the tenets are not binding department instructions, they are understood to be important standards for AI development.
What the Left Is Saying
Progressive critics and AI safety advocates argue that the lack of adherence to ethical guidelines poses significant risks to both military personnel and civilians. Researchers have long warned that AI weapons could become unpredictable or function with insufficient human control, potentially leading to unintended escalation of conflicts. The GAO findings highlight that six of the ten surveyed officials explicitly identified risks when using their projects in operational settings. Critics point to the understaffing of the Department of the Responsible AI Division as evidence that the administration is prioritizing speed over safety. With only three of ten authorized positions filled, advocates argue the Pentagon lacks the necessary expertise to ensure its AI systems are aligned with international norms and ethical standards.
What the Right Is Saying
Conservative voices and Defense Secretary Pete Hegseth frame the current approach as a necessary response to national security imperatives. Hegseth has repeatedly dismissed previous cautions regarding AI tools, arguing that bureaucratic delays and committee cycles measured in months or years are unacceptable in the context of modern warfare. In a speech at SpaceX in January, Hegseth stated that "military AI is going to be a race for the foreseeable future," arguing that the risks of moving too slowly outweigh the impacts of imperfect alignment. Supporters of this view contend that the Pentagon’s rapid adoption of autonomous technologies, such as the new warfare command focused on autonomous weapon development announced last month, is essential to maintain a technological edge over adversaries.
What the Numbers Show
The GAO survey involved 10 officials working on 10 distinct AI projects. Six of these officials reported that their teams had "identified risk(s)" with using their projects in an operational environment. Four officials stated that the requirements for their weapons and projects were not "aligned with Responsible AI tenets." Additionally, three of the ten officials told the GAO they did not know if their projects would be continuously monitored after deployment, despite the guidelines encouraging such monitoring. The Department of the Responsible AI Division was authorized to hire 10 staff members in 2023, but as of February, only three positions were filled. The GAO report states the Pentagon is "unlikely to fill vacant positions or increase staffing." The GAO issued two recommendations: clearer outlining of responsibilities for the AI oversight office and addressing understaffing. The Pentagon agreed with these recommendations but had not implemented changes as of Oct. 22.
The Bottom Line
The disconnect between the Pentagon’s "AI-first" rhetoric and the operational reality of its projects suggests a potential gap in risk management for autonomous weapons. While the Department has agreed to GAO recommendations to improve oversight and staffing, the lack of implementation leaves questions about how future AI deployments will be monitored. The GAO’s call for clearer responsibilities and meeting schedules for the AI oversight office remains pending. As the Pentagon continues to expand its AI portfolio, the extent to which its ethical guidelines are integrated into project lifecycles will be a key indicator of its commitment to responsible innovation versus rapid capability deployment. NPR did not receive a response from the Pentagon regarding when specific changes to address the GAO’s findings would be enacted.