The AI Threat Model

1. What are the three primary assets that need protection in an AI threat model?

Data, Model behavior, Audit trail Natural language, User intent, Model training User data, System performance, Output validity Application security, System design, User privacy

2. Which of the following represents a threat where attackers manipulate model behavior through hostile text?

Data exfiltration Prompt injection Denial of wallet Training-data poisoning

3. What is the main purpose of creating a written threat model in AI deployments?

To improve user experience To map recognized controls to assets and threats To enhance model training effectiveness To streamline data processing operations

4. How does LyraLearn address the threat of knowledge-base poisoning?

By using public datasets for training By allowing unrestricted data ingestion Through allowed content ingestion controls By focusing solely on user data

Continue: Prompt Injection Deep Dive → Review this lesson Retake quiz