Skip to content
AI Security

Interview question · How LLM Systems Work

What is the difference between abuse of an AI system and compromise of it?

Same product, two different adversaries and two different teams. The scored answer separates them by what the attacker gets, and shows where they meet.

mediumNewPreparation tags: OpenAI, Anthropic, MicrosoftUpdated Nov 2026

Same product, two different adversaries and two different teams. The scored answer separates them by what the attacker gets, and shows where they meet.

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 6 follow-ups like an interviewer would

A model provider sees one customer account generating phishing emails at scale and, separately, an attacker reading another customer's conversation history.

  1. How do abuse and compromise differ?
  2. Why does the distinction change the response?
  3. What if a jailbreak lets a user get the model to call a tool on another customer's data?
  4. What if the phishing account was opened with stolen identity details?
  5. How do the two teams work together?
  6. How would you organise detection and response?

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 trackBeyond prompt injection, what security problems come with accepting images, audio and documents?Next in this trackWhat changes for security when an application uses a reasoning model that thinks before it answers?Next in this trackDescribe a RAG pipeline as a data flow and mark every point an attacker can influence.
DISCUSSION · 0

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