Which elements are essential for mitigating data leakage and AI misuse in an AI security program?

• A. Occasional risk reviews and ad-hoc controls.
• B. Continuous risk management, monitoring, security policies, and controls.
• C. Only encryption at rest.
• D. Publicly sharing data to test risk.

Answer: (B)

Mitigating data leakage and AI misuse requires ongoing governance and layered safeguards that adapt as risks evolve. Continuous risk management provides a living view of where data leaks or model misuse could occur, tracking residual risks and prompting timely updates to people, processes, and technology as data, models, or regulatory requirements change. Continuous monitoring offers real-time visibility into data access patterns, training data provenance, model inputs and outputs, and potential exfiltration attempts or unusual model behavior, enabling rapid detection and response. Security policies establish clear rules for data handling, access, and sharing, and they translate into concrete controls such as least-privilege access, data classification and retention, encryption in transit and at rest, data loss prevention, and governance over model development and deployment. Together, these elements create defense in depth that scales with the AI lifecycle and evolving threats. Relying on one-off reviews, encryption alone, or unsafe practices like publicly sharing data to test risk fails to address the ongoing, dynamic nature of data protection and AI safety.