A municipal bus-shelter-ad tester is about to paste a complainant’s home address into a prompt to an LLM-powered test tool. Which skills should stop that action?
Select an answer to reveal the explanation.
Short Explanation
Think of the prompt as a postcard anyone on the route might read—strip or mask names and addresses first. Data sanitization and privacy-preserving prompts are core tester skills, plus knowing that sharing testware with an LLM has security weight. That is not a CT-AI SUT exercise.
Full Explanation
Testers must sanitize data by removing or masking sensitive, personal, or confidential information and practice privacy-preserving prompts; sharing testware with an LLM carries security implications. A municipal bus-shelter-ad tester about to paste a complainant’s home address into a prompt should stop and apply those skills. Naming the Chapter 3 attack vector for every possible model jailbreak is a different task from sanitization practice. Measuring metamorphic relations on the LLM as the product under test is CT-AI SUT work, not the Chapter 5 sanitization skill. Replacing all personal data with real production dumps for realism worsens the risk the skill is meant to prevent. Exam caveat: Chapter 5 frames sanitization and privacy-preserving prompts as required skills; naming Chapter 3 vectors or testing the model as SUT are different tasks. Operational check: mask or remove the complainant’s home address and other PII before any bus-shelter-ad prompt, and use only approved tools. Keep civic testware subject to the same review discipline you would apply to human-authored packs. On the written exam, prefer syllabus terminology over vendor marketing language or lab choreography. Treat fluent output as a draft until organizational quality gates and data-handling rules are satisfied.