Skip to content
AI Security

Interview question · LLM Application Security

Can prompt caching or response caching in an LLM application leak data between users?

Caches are shared state, and shared state crosses tenants. The scored answer separates response caches from prompt prefix caches and keys each so one user can never read or detect another's data.

mediumNewPreparation tags: OpenAI, Anthropic, GoogleUpdated Nov 2026

Caches are shared state, and shared state crosses tenants. The scored answer separates response caches from prompt prefix caches and keys each so one user can never read or detect another's data.

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

To cut cost, an LLM app caches responses and uses provider-side prompt caching for prompts shared across all customers.

  1. What if the response cache key is only the prompt text?
  2. What if a semantic cache returns answers for similar questions?
  3. What if prompt prefix caches are shared across tenants?
  4. What should never be cached across users?
  5. How would you test the cache design?
  6. What caching design would you use?

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 trackWhat security and privacy issues come with fine-tuning a model on customer data?Next in this trackWhy is overreliance on model output a security risk, and how do you mitigate it?Next in this trackSecure a chat-with-your-documents connector to a company document store, including permission sync.
DISCUSSION · 0

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