Skip to content
AI Security

Interview question · How LLM Systems Work

If the model is forced to return valid JSON matching a schema, is its output safe to use?

A schema can constrain fields and allowed values. Follow what remains: trusted facts, resource and task permissions, safe consumers and actual completion.

mediumNewUpdated Nov 2026

A schema can constrain fields and allowed values. Follow what remains: trusted facts, resource and task permissions, safe consumers and actual completion.

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.

10 answers per topic instead of 5, plus saved progress and bookmarks · costs nothingor unlock all 529 remaining answers · $25

Every worked answer in all 17 topics, every lesson in every course, the full concept curriculum, the complete practice-test bank, and the downloadable security field guides.

6 months · One payment · No auto-renewal

What if · the follow-ups in this answer

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

A fictional makerspace tool accepts a closed scheduling proposal with a registered operation, booking ID and one of two allowed slot labels. The model produces a proposal that meets those representation constraints.

  1. What if the proposal names another member’s booking using a valid identifier?
  2. What if the member owns both bookings but requested a change to only one?
  3. What if the schema enumerates the two allowed slots and the model chooses the unconfirmed one?
  4. What if validation passes and the booking becomes unavailable before the worker commits?
  5. How would you use structured output without overclaiming its guarantee?

Each rung has a nudge, a model answer, the principle that carries over and the weak answer to avoid. Premium opens every answer and ladder.

UP NEXT ON YOUR JOURNEY
Next in this trackHow would you use Google's Secure AI Framework (SAIF) and its risk map to choose controls for a new AI product?Next in this trackCan fine-tuning remove a model's safety behaviour? What does that mean for a product team?Next in this trackHow are spam, fraud and malware classifiers trained, and how do attackers evade them?
DISCUSSION · 0

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