Quick answer

Below are 10 original AIGP-style questions on the EU AI Act, each with a hidden answer and an explanation tied to the article it tests. They cover prohibited practices, high-risk classification, provider vs deployer duties, becoming a provider, transparency and deep fakes, general-purpose AI, FRIAs and penalties. Under BoK v2.1, AI-specific laws are about 6–8 scored questions, but the Act also runs through the development and deployment scenarios. Score yourself with the table at the end.

The EU AI Act is the part of the AIGP candidates over-read and under-practise. The exam doesn't ask you to recite it. It asks you to place a system in the right tier, work out who's the provider and who's the deployer, and pick the obligation that applies. These questions drill exactly that. See also how much of the AIGP is the EU AI Act.

How to use this set

  • Answer before you open the explanation. Commit to a letter and a reason. A right answer for the wrong reason counts as a miss.
  • Name the role and the tier first. Most EU AI Act questions are answered once you know who's the provider, who's the deployer, and which risk tier applies.
  • These are original questions written in the exam's style, not real exam items.

Practice questions

Q1Prohibited practices

A city government plans to score residents' "trustworthiness" from their social behaviour across many contexts and use the score to prioritise access to public services. Under the EU AI Act, this system is:

  • AHigh-risk, so it needs a conformity assessment
  • BA prohibited practice
  • CLimited-risk, so it only needs transparency notices
  • DOutside the Act, because it's a public-sector use
Show answer

Correct: B

Social scoring that evaluates people based on their social behaviour or personal characteristics, and leads to detrimental or disproportionate treatment, is one of the prohibited practices. It's banned outright, not managed through high-risk controls (A). Public-sector use (D) is no exemption; social scoring by public authorities was the original concern behind the ban.

EU AI Act, Art. 5(1)(c)
Q2Risk classification

A company uses an AI tool to rank job applicants' CVs and filter who gets an interview. How is this system classified?

  • AMinimal risk
  • BHigh-risk, as an Annex III employment use
  • CLimited risk, with transparency duties only
  • DProhibited
Show answer

Correct: B

AI used for recruitment or selection, including filtering applications and evaluating candidates, is listed in Annex III under employment and workers management, so it's high-risk. That triggers the full set of high-risk requirements for the provider and specific duties for the deployer. It isn't prohibited (D). The workplace prohibition targets emotion recognition, not CV ranking.

EU AI Act, Art. 6(2), Annex III(4)
Q3Roles · deployer duties

A retailer buys the CV-screening tool from Question 2 from a vendor and uses it internally, unchanged. Which obligation falls on the retailer?

  • ACarrying out the conformity assessment and affixing the CE marking
  • BDrawing up the technical documentation for the system
  • CAssigning human oversight to people with the necessary competence and authority
  • DSetting up the quality management system for the system's design
Show answer

Correct: C

The retailer is a deployer. Deployers must use the system according to its instructions, assign human oversight to competent people, monitor its operation, keep logs under their control, and inform workers before using high-risk AI in the workplace. Conformity assessment, CE marking, technical documentation and the quality management system (A, B, D) are provider obligations.

EU AI Act, Art. 26; Arts. 16–17
Q4Roles · becoming a provider

The retailer then retrains the vendor's CV-screening system on its own data, changing how it ranks candidates in a way the vendor never assessed, and markets it to other retailers under its own brand. What is the most important consequence?

  • ANone, because the vendor remains the provider
  • BThe system stops being high-risk
  • COnly the vendor's insurance needs updating
  • DThe retailer takes on the provider's obligations for the modified system
Show answer

Correct: D

A deployer or other third party that puts its name or trademark on a high-risk system, or makes a substantial modification to it, is treated as its provider and takes on provider obligations. Here both apply. The classification doesn't change (B); the obligations move.

EU AI Act, Art. 25
Q5Transparency

A bank launches a customer-service chatbot that answers account questions. Which obligation applies under the EU AI Act?

  • ANone, because chatbots are minimal risk
  • BThe chatbot must pass a third-party conformity assessment
  • CCustomers must give explicit consent before each chat
  • DUsers must be informed they're interacting with an AI system, unless that's obvious from the context
Show answer

Correct: D

AI systems intended to interact directly with people must be designed so those people are informed they're dealing with AI, unless that's obvious to a reasonably well-informed person. This is a transparency obligation, not a high-risk requirement (B), and it doesn't require consent (C).

EU AI Act, Art. 50(1)
Q6Transparency · deep fakes

A company uses a generative AI tool to create a realistic video in which its real CEO appears to announce a new product. The CEO never recorded it. The video will be published on social media. What does the Act require of the company as deployer?

  • ADisclose that the content has been artificially generated or manipulated
  • BNothing, because the CEO approved the use of their likeness
  • CRegister the video in the EU database for high-risk systems
  • DObtain approval from a market surveillance authority before publishing
Show answer

Correct: A

Deployers of AI systems that generate or manipulate image, audio or video content constituting a deep fake must disclose that the content is artificially generated or manipulated. Lighter rules apply to evidently artistic, creative or satirical work. A realistic video of a real person saying something they never recorded is exactly what the rule targets. The CEO's approval (B) doesn't remove the duty, because the disclosure protects the people watching.

EU AI Act, Art. 50(4)
Q7General-purpose AI

Which obligation applies to all providers of general-purpose AI models, not only those whose models pose systemic risk?

  • ACarrying out adversarial testing of the model
  • BReporting serious incidents to the AI Office
  • CPublishing a sufficiently detailed summary of the content used for training
  • DEnsuring an adequate level of cybersecurity protection for the model
Show answer

Correct: C

All GPAI model providers must keep technical documentation, provide information to downstream providers, have a copyright policy and publish a training-content summary. Adversarial testing, serious-incident reporting and cybersecurity protection (A, B, D) are additional obligations for models with systemic risk.

EU AI Act, Arts. 53 and 55
Q8Deployer duties · FRIA

A private bank plans to deploy an AI system that evaluates the creditworthiness of individual loan applicants. Before deployment, what must the bank carry out under the Act?

  • AA conformity assessment as provider
  • BA fundamental rights impact assessment
  • CA DPIA only. The AI Act adds nothing
  • DNothing, because credit scoring is limited-risk
Show answer

Correct: B

Deployers of high-risk systems for creditworthiness assessment of individuals (and for life and health insurance risk and pricing) must perform a fundamental rights impact assessment before first use, as must public bodies and private entities providing public services. A GDPR DPIA (C) may also be needed and can feed into it, but it doesn't replace the FRIA. See our FRIA guide.

EU AI Act, Art. 27; Annex III(5)(b)
Q9Prohibited practices · workplace

A call-centre operator wants to use AI to infer agents' emotions from their voices during calls to measure "engagement" for performance reviews. Under the Act this is:

  • APermitted, if agents are told about it
  • BHigh-risk, so it requires a FRIA
  • CProhibited, unless it's for medical or safety reasons
  • DLimited-risk, so it needs a transparency notice
Show answer

Correct: C

Placing on the market or using AI to infer emotions in the workplace (and in education) is prohibited, except where intended for medical or safety reasons. Performance management is neither. Telling the agents (A) doesn't lift a prohibition.

EU AI Act, Art. 5(1)(f)
Q10Penalties

What is the maximum fine for engaging in a prohibited AI practice?

  • A€35 million or 7% of worldwide annual turnover, whichever is higher
  • B€7.5 million or 1% of worldwide annual turnover
  • C€15 million or 3% of worldwide annual turnover
  • D€20 million or 4% of worldwide annual turnover
Show answer

Correct: A

Prohibited practices carry the top tier: up to €35 million or 7% of total worldwide annual turnover, whichever is higher. Most other operator obligations are capped at €15 million or 3% (C), and supplying incorrect information to authorities at €7.5 million or 1% (B). €20 million or 4% (D) is the GDPR's upper tier, a common distractor.

EU AI Act, Art. 99

How to read your score

ScoreWhat it means
9–10 with the right reasonsYou've got the Act's structure. Move on to mixed scenario practice across all four domains
6–8Typical a few weeks out. Your misses show you exactly what to review
5 or fewerRebuild your map of the Act (tiers, roles, obligations) before more questions. See risk classifications with real examples
A note on dates: These questions test the Act's substance, not its application dates. Phase-in deadlines have been the subject of 2026 amendment proposals, so check the BoK and current sources before relying on a specific date.

What your misses usually mean

  • Missed Q3 or Q4? Roles are your gap. The exam loves the provider/deployer split and the ways a deployer becomes a provider.
  • Missed Q1 or Q9? Learn the prohibited list as a list. These are reliable marks once memorised.
  • Missed Q7? Separate the baseline GPAI duties from the systemic-risk extras.

300 more like these, across all four domains

The AIGP Complete Pack: 300 original questions in all six formats, mapped to BoK v2.1, with every option explained, plus a 100-question timed mock, a Cram Guide with a five-scenario masterclass, and an After You Pass guide.

Get the Complete Pack

Frequently asked questions

How many EU AI Act questions are on the AIGP exam?

Under BoK v2.1, AI-specific laws (the EU AI Act plus other AI laws) are about 6–8 scored questions. The Act also appears indirectly in development and deployment scenarios, so it matters more than that number suggests.

Do I need to memorise EU AI Act article numbers for the AIGP?

No. Know the risk tiers, the roles (provider, deployer and others), who owes which obligations, and the transparency rules. Article numbers only help you organise your notes.

Are these real AIGP exam questions?

No. They're original questions written in the exam's style. The IAPP doesn't release live exam items.

Sources

  1. EUR-Lex: Regulation (EU) 2024/1689 (Artificial Intelligence Act)
  2. IAPP: AIGP certification and Body of Knowledge

Last reviewed: November 2026.