Skip to content
AI Security

Interview question · Software, Data & Model Supply Chain

How do you verify the integrity and lineage of web-scraped training data?

Scraped data changes under you and anyone can write to the web. The scored answer snapshots and hashes what was collected, records where it came from, filters what it can, and keeps enough lineage to undo a mistake.

hardNewPreparation tags: Google DeepMind, Meta, OpenAIUpdated Nov 2026

Scraped data changes under you and anyone can write to the web. The scored answer snapshots and hashes what was collected, records where it came from, filters what it can, and keeps enough lineage to undo a mistake.

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 training dataset is assembled from web crawls, vendor data and internal logs, and stored as a list of URLs plus processing scripts.

  1. Why is a URL list without hashes a risk?
  2. What should be recorded at collection time?
  3. How do you track processing?
  4. How should sources be treated differently?
  5. What if a model behaviour is later traced to bad data?
  6. What provenance system 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 trackYour AI vendor announces a breach. What do you check in your own environment?Next in this trackWhat are reproducible builds, why do they matter for supply chain security, and can ML training be reproducible?Next in this trackWhat security tests would you run before every model update ships?
DISCUSSION · 0

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