Recognition Before Authority: What AI Should and Should Not Do in the First Minutes of a Working Incident

Recognition Before Authority examines what AI should and should not do in the first minutes of a working incident. It argues AI should improve the commander’s picture—not take strategy, interior commitment, or withdrawal. The paper sets a Display → Recommend → Act model, keeps known/inferred/unknown/unsupported separate, and maps near-term use to Display and bounded Recommend only. It reviews NIOSH, NIST SP 1500-29, and fire-service adoption evidence, states that no field trial yet proves safety outcomes, and offers a pilot scoreboard before any increase in authority. For fire and public-safety leaders.