AI Risks in Cybersecurity: Review the Action Before Automating
AI can help summarise security information, but its mistakes become more consequential when its output triggers actions. Review the evidence, affected systems, permissions and recovery plan before connecting recommendations to blocking, account changes or other operational controls. Separate ordinary model error, sensitive-data exposure and deliberate manipulation of an AI system.