AISP practice

25 questions. Choose the BEST answer.

You can answer as a guest, but you'll need a free account to see your score and save it. Sign up · Sign in
Question 1 Question 1 · M2
Why is an embedding index of HR documents treated as personal data?
Question 2 Question 2 · M3
A team replaces customer names with reversible tokens before fine-tuning and calls the data anonymous. Is that correct?
Question 3 Question 3 · M4
Where should policy enforcement points be placed in an AI system?
Question 4 Question 4 · M5
A guardrail fix for a jailbreak fails once when the tester retries the original prompt. Can the ticket be closed?
Question 5 Question 5 · M3
Which measure reduces both training-data extraction and membership inference risk?
Question 6 Question 6 · M3
A team wants to rely on cryptographic erase for a decommissioned training cluster. What must be true?
Question 7 Question 7 · M6
An open-weight model is labeled with a permissive license but was fine-tuned from a base model with a non-commercial license. Can it be used commercially?
Question 8 Question 8 · M1
A system prompt says 'never reveal personal data', yet the model sometimes does. What principle explains this?
Question 9 Question 9 · M1
An HR chatbot uses RAG over the HR library and answers a contractor's question with a manager's salary from a spreadsheet in the same library. Which single control would have prevented this?
Question 10 Question 10 · M5
Why should legal and privacy join an AI incident early?
Question 11 Question 11 · M1
Reviewers override a loan model's recommendations in 0.1% of cases, against 6% during testing. What is the MOST likely concern?
Question 12 Question 12 · M6
An attacker floods a public assistant with prompts designed to maximize output length. Which OWASP LLM risk is this?
Question 13 Question 13 · M1
An investigator re-asks a chatbot the question that produced a harmful answer and gets a safe reply. What can be concluded?
Question 14 Question 14 · M3
A data scientist downloads a pickle-format model from a public hub to test it. What is the main risk?
Question 15 Question 15 · M5
Investigators need to know whether a RAG assistant's permission check ran for a request that leaked another client's data. Which log is essential?
Question 16 Question 16 · M5
A unique string planted in a chatbot's system prompt appears in a post on a public forum. What does this MOST likely indicate?
Question 17 Question 17 · M6
A vendor claims its model is 'robust to adversarial inputs'. What should you request?
Question 18 Question 18 · M5
Poisoned data was used to fine-tune a model, which was then distilled into two smaller models. What must eradication include?
Question 19 Question 19 · M6
A red-team report states only 'the jailbreak succeeded'. What is missing?
Question 20 Question 20 · M2
An AI awareness program reports 98% course completion. What is the BEST next measure of effectiveness?
Question 21 Question 21 · M1
Why should an AI team log the exact retrieved chunks and prompt version for each response?
Question 22 Question 22 · M2
An AI vendor contract allows 7 days to notify incidents. Your organization must notify CERT-In within 6 hours. What should you do?
Question 23 Question 23 · M4
How can a pipeline enforce that a model was evaluated before deployment, rather than relying on a checkbox?
Question 24 Question 24 · M4
Which guardrail layer decides whether a successful prompt injection achieves anything harmful?
Question 25 Question 25 · M1
A model's inputs look the same as at launch, but the relationship between inputs and the correct outcome has changed after a regulation change. What type of drift is this?