Section Quiz — The AI Threat Landscape

Q1. Which of the following is NOT a stage in the AI lifecycle threat map?

Q2. A fraud model begins approving transactions it previously blocked. The most likely first-step diagnosis is:

Q3. Why do traditional security tools struggle to detect adversarial AI inputs?

Q4. According to IBM's Cost of a Data Breach Report 2026, what proportion of malicious breaches are now AI-enabled?

Q5. A model's outputs are wrong purely because the world changed and it wasn't retrained — with no malicious intent. In this lesson's taxonomy, this is:

Q6. Why can a standard WAF (Web Application Firewall) not stop an adversarial example crafted to fool a vision model?