Skip to content
AI Security

Interview question · AI Red Teaming, Evaluation & Response

How do you assess the severity of a vulnerability in an LLM application?

Rate the supported consequence and conditions, explain retry assumptions and separate severity from deployment risk and response priority.

mediumNewUpdated Nov 2026

Rate the supported consequence and conditions, explain retry assumptions and separate severity from deployment risk and response priority.

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 4 follow-ups like an interviewer would

A support-assistant test confirms one forbidden synthetic-note release. The wider scope is unknown, and a separate selected test set has a reported 5% success rate.

  1. What if someone calls success nearly certain within twenty retries?
  2. What if the output has no tool side effect?
  3. What if containment is available before the final rating is agreed?
  4. How should the report support a defensible release decision?

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 trackWhy can a single evaluation number mislead you about an AI system's security?Next in this trackWhat is over-refusal, and why does it belong in a security evaluation?Next in this trackWhat do public jailbreak and prompt injection benchmarks measure, and what are their limits?
DISCUSSION · 0

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