Skip to content
AI Security

Interview question · How LLM Systems Work

Explain to a non-technical manager why you cannot simply delete sensitive data from a trained model.

A communication question with a technical core. The answer that scores uses one good analogy, stays accurate, and ends with options and a decision.

mediumNewUpdated Nov 2026

A communication question with a technical core. The answer that scores uses one good analogy, stays accurate, and ends with options and a decision.

Make your first attempt

What would you say before reading the answer?

  1. 01 State the assumptions your answer depends on.
  2. 02 Explain the mechanism and choose a control.
  3. 03 Name a condition that would change your decision.

The worked answer includes an approach, scoring criteria and follow-up questions. Compare the reasoning with your own.

signing in unlocks more free answers per topic

What if · the follow-ups in this answer

The answer works through 5 follow-ups like an interviewer would

A fictional makerspace member withdraws a support transcript. It appeared in dataset D1, which trained adapter A1 on base B0. The team also merged A1 into model M1 and kept M1 as a fallback.

  1. What if the team deletes the transcript from D1 and keeps serving A1?
  2. What if the team rebuilds A1 from B0 without the transcript?
  3. What if A1 is disabled but the merged M1 remains the outage fallback?
  4. What if a vendor reports zero reproductions after an unlearning procedure?
  5. How would you explain the next step to the manager?

Each rung has a nudge, a model answer, the principle that carries over and the weak answer to avoid. Sign in free to open it.

UP NEXT ON YOUR JOURNEY
Next in this trackWhat is safety training, and why does a red teamer care how a model was aligned?Next in this trackWhat does the OWASP Top 10 for Agentic Applications add to the LLM Top 10?Next in this trackHow does NIST AI 100-2 classify adversarial machine learning attacks, and why does a shared taxonomy help?
DISCUSSION · 0

No comments yet. Share a correction or a different way to answer it.