Artificial Intelligence Governance Professional (AIGP)

Last checked November 10, 2026 against the official IAPP exam page.

1. Question 1Domain III - Understanding how to govern AI development

   A team fine-tunes a model on a dataset obtained from a third party. The original source, the licence terms and the transformations applied to the data were never recorded. Which governance gap matters most in this situation?

   1. AMissing data lineage and provenance records, so the source and permitted use of the training data cannot be shown
   2. BInsufficient hyperparameter tuning, which means the fine-tuned model may not have reached its best performance on the task
   3. CToo many model versions stored in the registry, which makes it harder for developers to find the latest production build
   4. DMissing user acceptance testing for the interface, which means business users have not yet approved the final presentation
   Show answer and explanationHide answer

   Correct answer: A

   Without lineage and provenance, the organization cannot verify its rights to use the data, trace errors back to their source or respond to audits. The other gaps are real quality issues, but they do not create the legal and accountability exposure described in the scenario.

2. Question 2Domain IV - Understanding how to govern AI deployment and use

   A government agency plans to use a general-purpose language model inside a benefits eligibility tool that decides which residents receive assistance. Which deployment risk factor should most strongly shape the governance approach?

   1. AThe logo and brand reputation of the model vendor, since well-known vendors generally carry lower operational risk
   2. BThe severity of consequences for individuals if the tool errs, since the decision affects their access to assistance
   3. CThe number of training tokens used during the model's pretraining, which determines how accurate its outputs will be
   4. DThe programming language used by the front-end team, which determines how easily the interface can be maintained
   Show answer and explanationHide answer

   Correct answer: B

   The impact of errors on individuals in the decision context is the key deployment risk factor, and it determines how strict oversight, testing and appeal rights must be. Vendor brand, pretraining volume and front-end language do not measure the harm a wrong decision would cause.

3. Question 3Domain II - Understanding how laws, standards and frameworks apply to AI

   A company collected customer support transcripts to resolve tickets. It now wants to reuse them to train a new AI model for marketing analytics. Which data protection principle is most directly engaged?

   1. AData minimization, which requires limiting the amount of personal data collected to what is needed for the purpose
   2. BPurpose limitation, which requires that later uses be compatible with the purposes for which the data was originally collected
   3. CStorage limitation, which requires that personal data be kept no longer than necessary for the stated purposes
   4. DAccuracy, which requires that personal data be kept accurate and, where necessary, kept up to date so that decisions rely on correct information
   Show answer and explanationHide answer

   Correct answer: B

   Reusing data for a new purpose such as model training requires a compatibility assessment or a new lawful basis under purpose limitation. Minimization, storage limitation and accuracy are real principles, but none of them governs whether a new purpose is permitted.

4. Question 4Domain I - Understanding the foundations of AI governance

   Which statement best distinguishes machine learning from a rule-based expert system?

   1. AMachine learning systems are always more transparent than rule-based systems because their decision logic is written in plain language
   2. BRule-based systems learn continuously from production data, whereas machine learning systems must be reprogrammed by hand for each update
   3. CRule-based systems cannot be used for decisions that affect individuals, whereas machine learning systems can be used for those decisions
   4. DMachine learning systems derive patterns from data to produce outputs, while rule-based systems follow logic that people have written explicitly
   Show answer and explanationHide answer

   Correct answer: D

   Machine learning infers patterns from training data, while rule-based systems apply if-then logic that developers encode directly. The other options are wrong: learned models are often harder to interpret than explicit rules, rule-based systems are routinely used for individual decisions, and the description of learning and reprogramming is reversed.

5. Question 5Domain III - Understanding how to govern AI development

   A lending model shows similar overall accuracy for all applicant groups. However, the approval rate for one protected group is only 60 percent of the approval rate for the reference group. What is the most important conclusion?

   1. AThe selection rate ratio falls below the four-fifths benchmark, which signals possible adverse impact that needs investigation
   2. BNo concern exists because overall accuracy is similar across groups, and accuracy is the primary measure of model quality
   3. CThe approval gap is caused by different model latency for each group, so the team should focus on infrastructure tuning
   4. DThe validation loss is lower than the training loss, which shows the model has generalized well to all applicant groups
   Show answer and explanationHide answer

   Correct answer: A

   A selection rate ratio below 80 percent, the four-fifths guideline, is a standard signal of potential adverse impact that requires investigation. Similar overall accuracy does not rule out unequal outcomes, and latency and loss comparisons say nothing about the approval disparity.