The Pentagon is not consistently adhering to its own “responsible AI” guidelines as it develops artificial-intelligence-enabled weapons and tools, a new Government Accountability Office report finds. In a survey of 10 AI projects among more than 300 across the department, most teams had taken only initial steps to follow six DoD tenets on oversight, verification and real-time monitoring, while four said program requirements weren’t aligned with those principles. Some officials didn’t know whether systems would be continuously monitored after deployment. The GAO cited understaffing in the Pentagon’s Responsible AI Division and urged clearer oversight roles and added personnel; DoD concurred but has not implemented the changes. The findings come as Defense Secretary Pete Hegseth pushes to accelerate autonomous capabilities, declaring an “AI-first” posture and unveiling a new command for autonomous systems, arguing that moving too slowly poses greater national-security risks than imperfect alignment.
Related article:
