Skip to content
AI Security

Interview question · Adversarial ML & Model Privacy

How are backdoors implanted through fine-tuning, and why are they hard to find?

The model passes your evaluation because your evaluation does not contain the trigger. And safety training may teach it to hide better.

hardNewPreparation tags: Anthropic, Google DeepMind, OpenAIUpdated Nov 2026

The model passes your evaluation because your evaluation does not contain the trigger. And safety training may teach it to hide better.

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.

signing in unlocks more free answers per topic

What if · the follow-ups in this answer

The answer works through 6 follow-ups like an interviewer would

Your team plans to deploy a third-party fine-tuned model from a public hub for internal code generation. Research shows models can be trained to behave normally until a trigger appears, even after safety training.

  1. How does a backdoor get into a model?
  2. Why are backdoors hard to find?
  3. What if the model's publisher is unknown?
  4. How could a backdoor harm the code generation use case?
  5. Can probing or interpretability tools clear the model?
  6. What would you require before deploying the model?

Each rung has a nudge, a model answer, the principle that carries over and the weak answer to avoid. Sign in free to open it.

UP NEXT ON YOUR JOURNEY
Next in this trackWhy does exposing confidence scores make extraction and inference attacks easier?Next in this trackAccuracy drops 20 percent. How do you tell an attack from drift?Next in this trackExplain FGSM and PGD, and what the perturbation budget means.
DISCUSSION · 0

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