AI Targeting Systems Can Confidently Pick Wrong Targets

A former Pentagon AI specialist is cautioning that military artificial intelligence systems can present convincing arguments for striking incorrect targets, and service members need training to identify and push back on those recommendations. The warning goes beyond technical glitches to highlight how AI can generate what appears to be solid reasoning while leading to fundamentally flawed decisions in combat situations.

Rather than announcing new rules or immediate changes, the message serves as a broader alert about potential failure modes in AI-assisted targeting, surveillance, and mission planning. As artificial intelligence becomes more integrated into military operations, troops and decision-makers will need to develop skills to recognize when AI confidence actually masks underlying errors in judgment or analysis.