Interview Prep
Responsible AI Interview Questions
Bias, privacy, prompt injection and safe deployment.
16
Intermediate
What is prompt injection?
When untrusted input (a web page, an email, a document) contains instructions that hijack the model's behaviour. Mitigate by separating instructions from data, limiting tool permissions and reviewing high-impact actions.
Key Points:
- Treat retrieved content as data, never as instructions.
17
Intermediate
How do you protect PII in an LLM app?
Minimise what you send, redact or mask PII before calling the model, use providers with suitable data-retention terms, and log carefully.
Key Points:
- Least data, least privilege.
18
Beginner
How can bias show up in AI systems?
Through skewed training data, labels, or how outputs are used. Test outputs across groups and keep humans in the loop for consequential decisions.
Key Points:
- Measure, don't assume.