Skip to content
AI Security

Interview question · Software, Data & Model Supply Chain

How would you check a public pre-trained model for a backdoor, and what can you not detect?

You can check the file and test the behaviour; you cannot prove a negative. The scored answer separates code-level threats from learned backdoors and is honest about what testing can find.

hardNewPreparation tags: Google DeepMind, Microsoft, MetaUpdated Nov 2026

You can check the file and test the behaviour; you cannot prove a negative. The scored answer separates code-level threats from learned backdoors and is honest about what testing can find.

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 team wants to fine-tune an open-weight model downloaded from a public hub for an internal assistant that can query company databases.

  1. How do you check whether the file is safe to load?
  2. How do you look for a behavioural backdoor?
  3. Can testing prove there is no backdoor?
  4. Will your own fine-tuning remove a backdoor?
  5. What system controls limit the damage if it is backdoored?
  6. How would you approve the model?

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 trackHarden a CI/CD pipeline: runners, secrets, pinned actions and branch protections.Next in this trackWho should be allowed to promote a model in the registry, and how do you detect tampering?Next in this trackWhat scans and admission policies would you require before an open-weights model runs in production?
DISCUSSION · 0

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