Skip to content
AI Security

Interview question · LLM Application Security

Design an ingestion pipeline for a RAG index that resists poisoning, and explain how you would estimate the poisoning risk that remains.

Research has shown a handful of planted documents can steer targeted answers. The defence is not a detector that recognises poison; it is control over who can write, staged promotion, and a risk estimate that multiplies the stages.

hardNewPreparation tags: Microsoft, Google, DatabricksUpdated Nov 2026

Research has shown a handful of planted documents can steer targeted answers. The defence is not a detector that recognises poison; it is control over who can write, staged promotion, and a risk estimate that multiplies the stages.

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 RAG system ingests documents from internal wikis, shared drives and public websites to answer employee questions.

  1. What if anyone can edit the wiki pages that are ingested?
  2. What checks belong in the ingestion pipeline?
  3. What if public websites are ingested?
  4. What if retrieved content carries trust labels?
  5. What if a poisoned document is discovered?
  6. What ingestion pipeline 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 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.Next in this trackYour CISO wants to buy an AI security platform. How do you evaluate one, module by module, and what will it never do for you?
DISCUSSION · 0

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