PENTAGON quietly revised US military targeting guidance in April, approving
language that envisions "systems where AI, under human supervision, can initiate
actions," a shift from the current human-in-the-loop requirement that actions be
initiated by humans. The revision underscores the Pentagon's push to accelerate
AI adoption; rapid advances in AI capabilities also introduce new threats and
ethical and legal dilemmas that could complicate compliance with the law of
armed conflict.
