Fire/EMS leadership sees an AI assistant invent hydrant locations during drills. Which action best treats that integrity failure in municipal incident taxonomy?

An AI assistant inventing hydrant locations during Fire/EMS drills is an integrity failure that can misdirect crews in a real response. Treating AI integrity failures that invent operational facts as reportable events in the critical-incident taxonomy ensures such hallucinations are logged, investigated, and corrected rather than dismissed as expected model quirks. Adding AI integrity failures that invent operational facts to the GIS data-quality taxonomy as routine map reviews demotes an AI integrity failure to routine map maintenance, keeping it out of the critical-incident channel where drill leadership and after-action review can see it. Adding AI integrity failures that invent operational facts to the vendor management taxonomy as service improvement items delays correction to a future service improvement cycle instead of treating the current municipal safety failure as a reportable incident. Adding AI integrity failures that invent operational facts to the exercise taxonomy as planned injects for evaluation hides the AI integrity failure inside exercise scoring instead of logging it as a reportable critical incident, leaving the underlying defect uncorrected. Exam caveat: not every minor wording glitch belongs in the critical taxonomy—criteria should target invented operational facts that could affect response, routing, or resource decisions. Operational check: add a drill inject where the assistant returns a false hydrant location and confirm participants open a reportable AI integrity incident under the updated taxonomy.