What is the purpose of AI threat modeling in SecAI+ and which lifecycle stages should it cover?

• A. It only assesses model accuracy and latency for deployment decisions.
• B. It focuses on optimizing data labeling throughput and annotation speed.
• C. It identifies security and privacy risks in AI systems and should cover data collection, training, deployment, monitoring, and feedback loops, as well as governance and compliance.
• D. It analyzes user behavior to drive product marketing strategies.

Answer: (C)

Threat modeling in SecAI+ is about identifying and mitigating security and privacy risks that arise throughout an AI system’s life cycle. It looks at how data moves and transforms from collection and labeling through preprocessing, training, and deployment, then continues with monitoring and feedback loops as the model is updated. By mapping out threat surfaces at each stage, you can spot where data could be exposed, manipulated, or misused—such as data poisoning during training, leakage through model outputs, or privacy risks from membership inference. It also encompasses governance and compliance, ensuring proper data provenance, access controls, auditing, and accountability across changes and iterations.

This makes the comprehensive view that covers data collection, training, deployment, monitoring, feedback loops, and governance/compliance the best fit. The other options focus on performance metrics, labeling throughput, or marketing analytics, which don’t address security and privacy risks across the AI life cycle.