Skip to content
AI Security

Interview question · Prompt Injection & Jailbreaks

Do delimiters, spotlighting and an instruction hierarchy help against prompt injection? What can you honestly claim for them?

Separate source cues, message integrity, learned instruction priority and service permissions. Compare the actual configurations on both violations and useful manual-reading work.

mediumNewUpdated Nov 2026

Separate source cues, message integrity, learned instruction priority and service permissions. Compare the actual configurations on both violations and useful manual-reading work.

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 4 follow-ups like an interviewer would

Eva reviews a fictional manual-answering assistant. Suppliers edit source documents, while the application controls workspace grants and the answer audience. The team proposes delimiters, spotlighting and hierarchy training.

  1. A manual includes text that resembles a system-role label. Has it changed the real API role?
  2. A transformed manual is always refused, including ordinary troubleshooting questions. Did the defense succeed?
  3. A team copies the StruQ front-end format into an unrelated model. Can it claim the paper's result?
  4. What would you put in the review record?

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 trackHow has the move from chatbots to agents changed the impact of prompt injection?Next in this trackHow do you evaluate many-shot and multi-turn jailbreak defenses?Next in this trackWhat did universal adversarial suffixes show about the robustness of aligned language models?
DISCUSSION · 0

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