Maintain current knowledge of how AI systems can be attacked, misused or behave unexpectedly, including prompt injection, jailbreaking, data poisoning, model theft, hallucination and unsafe autonomous actions, and the preventive and detective controls, such as guardrails, output filtering, monitoring, human-in-the-loop oversight and access controls on agentic actions, used to manage those risks. Cloud Security, Cybersecurity, Data Integrity, Identity Access Management (IAM), Risk Assessments, Risk Management, Risk Management Framework, Security Architecture Design, Security Compliance, Security Policies, Security Risk, Third Party Risk Management, Vulnerability Management, Web Application Security.