City security operations wants the information security program to cover misuse of municipal generative assistants. Which program action best expands detection coverage?

City security operations expands coverage of municipal generative-assistant misuse by adding AI-specific detection capabilities—such as prompt-firewall or abuse alerts—to the program’s detection use-case catalog. Generative misuse (sensitive data exfiltration via prompts, jailbreaks, toxic or fraudulent outputs) will not reliably surface in traditional malware or perimeter detections; cataloged AI use cases make monitoring intentional and testable. Removing generative assistants from monitoring scope to reduce alert volume shrinks visibility where new abuse is likely. Relying solely on end-user honor systems without detection use cases abandons SOC assurance. Documenting detections only after a public scandal forces retrospective logging is reactive program management. Exam caveat: AI detections need tuned baselines and privacy-aware log handling so prompt monitoring itself does not become an uncontrolled sensitive-data store. Operational check: add prompt-abuse and sensitive-entity-in-prompt detections to the use-case catalog with owners and runbooks, enable alerts in the municipal assistant logging pipeline, and conduct a tabletop that exercises escalation from an AI abuse alert to incident response.