Skip to content
AI Security

Interview question · LLM Application Security

Cosine similarity decides what a RAG system retrieves. How would you make retrieval prefer trustworthy sources instead of just similar ones?

The most similar chunk is often the one written to be retrieved. The answer adds provenance to every chunk, ranks with it outside the model, and refuses to let text declare its own authority.

hardNewPreparation tags: Microsoft, Google, AnthropicUpdated Nov 2026

The most similar chunk is often the one written to be retrieved. The answer adds provenance to every chunk, ranks with it outside the model, and refuses to let text declare its own authority.

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 retrieval system mixes results from vetted policy documents and user-submitted forum posts, and the assistant can file tickets.

  1. What if a forum post contradicts the official policy?
  2. What if untrusted results are in context when the model can take actions?
  3. How do you represent trust to the model?
  4. How does the user know where an answer came from?
  5. What if a trusted source is compromised?
  6. What would you build?

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 trackDesign an ingestion pipeline for a RAG index that resists poisoning, and explain how you would estimate the poisoning risk that remains.Next in this trackArchitect the guardrail pipeline for a customer-facing financial assistant, including how you guard streamed responses.Next in this trackPractical coding: implement a rate limiter for an LLM API that budgets tokens, not requests, then make it work across many gateway nodes.
DISCUSSION · 0

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