Certifications AIGP CIPP/E CIPP/US CIPM CIPT All certifications Career plan How it works Pricing Log in Start free
Study guide

The complete AIGP study guide

What the exam asks, in what order to study it, how long it takes, what it costs and where candidates lose marks.

AI-generated image - a professional with a folder outside a European institution, for the CIPP/E study guide
What it is

What is the AIGP?

The AIGP, the Artificial Intelligence Governance Professional, is the certification for people who govern how AI systems are built, bought and used. It tests whether you can apply the law, the standards and the frameworks to a real system and a real organisation, not only whether you can recite them.

It is issued by the IAPP. There are no prerequisites: no degree and no minimum experience. You sit it online or at a test centre, in English.

Format and fees

100questions85 scored, 15 unscored
165minutesoptional 15-minute break
300of 500 to passscaled score
$799exam feeretake $625
12monthsto sit after you buy
20CPE creditsevery two years, plus the fee

Membership of the IAPP costs $295 a year and includes the maintenance fee; without membership that fee is $250 every two years. Members pay $649 for the exam and $475 for a retake.

Before you plan anything, find out what you already know. Free account, fifteen minutes.Free AIGP assessment
The body of knowledge

The four domains, and what each one asks

The IAPP publishes a body of knowledge for each certification: the list of competencies the exam is written from, with the minimum and maximum number of questions per domain. The current version for AIGP is 2.1, in force since 2 February 2026. The 22Academy course follows it concept by concept.

1

Understanding the foundations of AI governance

16 to 20scored questions

What AI is and why it needs governing: how the systems work, where the risks and harms come from, the principles of responsible AI, and how an organisation sets up roles, policies and a governance programme.

Where people lose marks

The principles. Transparency, explainability and accountability are three different things, and the exam offers each as the plausible wrong answer for the others.

From the 22Academy bank · definition

Which principle of responsible AI requires that a named person or function can be held to answer for the outcomes of an AI system?

Twenty more like it, free
2

Understanding how laws, standards and frameworks apply to AI

21 to 25scored questions

The rules: the EU AI Act with its risk categories and its roles, the laws that already applied before it, such as data protection, and the frameworks organisations work with, among them the NIST AI Risk Management Framework and ISO/IEC 42001.

Where people lose marks

The role. Obligations under the EU AI Act follow from being a provider or a deployer, and candidates answer for the wrong one.

From the 22Academy bank · two-statement true or false

Statement I: under the EU AI Act, every AI system used in recruitment is prohibited. Statement II: a deployer of a high-risk AI system must assign human oversight to people with the competence, training and authority to carry it out. Which statement is correct?

Twenty more like it, free
Screenshot of the 22Academy course - accuracy per domain and module, with the weakest module marked as the one to focus on now
Screenshot from results in the course
Domain by domain

Three domains weigh the same. Which one is yours?

Domains 2, 3 and 4 each carry up to 25 of the 85 scored questions, so there is no single domain to lean on. The course shows your accuracy per domain and per module, and points at the module to work on now.

3

Understanding how to govern AI development

21 to 25scored questions

Governance while the system is being built: defining the use case, governing the data, training and testing, documentation, and the decision to release.

Where people lose marks

The documents. A model card, technical documentation and a data sheet answer different questions for different readers, and the exam asks which one fits.

From the 22Academy bank · scenario

A team is about to train a credit-scoring model and finds that one region is barely represented in the training data. What should it do before training continues?

Twenty more like it, free
4

Understanding how to govern AI deployment and use

21 to 25scored questions

Governance once the system is in use: the decision to deploy, assessing vendors, monitoring in production, handling incidents and taking a system out of service.

Where people lose marks

Responsibility after go-live. Buying a system from a vendor does not move the deployer's duties to the vendor.

From the 22Academy bank · best answer

A hospital uses a triage system from a vendor and notices that its accuracy is falling for older patients. What is the BEST first step for the hospital as deployer?

Twenty more like it, free

Not using your IAPP membership? Pay only the $250 maintenance fee and keep $340 every two years.

Start AIGP free
The rules

The articles

The EU AI Act comes back in every domain. Know these four groups of articles and the scenario questions become readable.

Prohibited practices

Article 5

The uses of AI that are banned outright, such as social scoring and manipulative techniques.

High-risk classification

Article 6 and Annex III

When a system is high-risk: a safety component of a regulated product, or a use on the list.

Requirements for high-risk systems

Articles 9 to 15

Risk management, data governance, technical documentation, record-keeping, human oversight, accuracy and robustness.

Providers and deployers

Articles 16 to 27

What each role must do, including the fundamental rights impact assessment for certain deployers.

Transparency

Article 50

Telling people they are dealing with an AI system, and marking generated content.

General-purpose AI models

Articles 51 to 55

The duties of model providers, and the extra duties for models with systemic risk.

Worth knowing

The guidelines

The body of knowledge names frameworks and standards next to the law, and they are examinable. Knowing what each one is for matters more than memorising its clauses.

NIST AI Risk Management Framework

Four functions: Govern, Map, Measure and Manage. Governance runs across the other three.

ISO/IEC 42001

The international standard for an AI management system, and the one an organisation can be certified against.

OECD AI Principles

The international baseline for trustworthy AI, and the source of the definition of an AI system used in law.

UNESCO Recommendation on the Ethics of AI

A global ethical framework, adopted by the UNESCO member states.

Ethics Guidelines for Trustworthy AI

The seven requirements of the EU High-Level Expert Group, the forerunner of the EU AI Act.

NIST Generative AI Profile

The risks specific to generative AI, mapped to the four functions.

Decisions

Case law

AI has little case law of its own yet. These decisions show how courts apply existing law to automated systems.

SCHUFA, Court of Justice of the EU, 2023

A credit score is itself an automated decision under Article 22 of the GDPR when the lender relies heavily on it.

Dun and Bradstreet Austria, Court of Justice of the EU, 2025

A person is entitled to a real explanation of the logic behind an automated decision about them.

SyRI, District Court of The Hague, 2020

A welfare fraud risk-scoring system was struck down for lack of transparency, under the right to private life.

State v. Loomis, Wisconsin Supreme Court, 2016

A court may use a proprietary risk score in sentencing, but only with warnings about its limits.

Mata v. Avianca, US District Court, 2023

Lawyers were sanctioned for filing case citations that a chatbot had invented.

Moffatt v. Air Canada, British Columbia tribunal, 2024

An airline was held to the wrong information its own chatbot gave a customer.

Always current

When the rules change, your course changes with them.

Every explanation in the course is stored once and linked to its source. When a regulation, a guideline or a decision changes, we update it in one place and every lesson that uses it is correct from that moment.

AI-generated image - a hand adds a new volume to a row of law books, as the course is updated when the rules change
The questions

How the questions are built, and how to read them

Exam questions come in a small number of structures, and a candidate who knows the content can still lose marks on the structure. The course tags every question with its structure; the technique is one tap away on each one, and every test shows whether you lost the mark on the content or on the type of question.

Two-statement true or false
Judge each statement on its own before you look at the options.
Best, most, primary
Underline the qualifier. Several options are defensible; one is the most complete.
Except, least, not
Three options are right. You are looking for the one that is wrong.
Only if, unless, provided that
Restate the rule without the condition, then test whether the outcome could happen without it.
Scenario
Read the question first, then the scenario. The scenario carries deliberate noise.
Definitions
The trap is a neighbouring term with one word changed.
Screenshot of the 22Academy course - score per question type, showing which type costs you the most marks
Screenshot from results in the course

Buy the exam voucher when you are ready, not before. It expires a year after you buy it.

Free AIGP assessment
How long it takes

How long does it take to prepare?

The IAPP recommends at least 30 hours of study. People who already work in privacy or with AI systems need less; people new to both need more. At three hours a day you can be ready in about five weeks.

Book the exam before you feel ready: a fixed date concentrates the effort. You have twelve months from purchase to sit it.

AIGP course€37 a week

Or €117 a month with a trial exam included.

Start the course

A five-week plan, at three hours a day

  1. Week 1The free assessment and domain 1: what AI is, the principles and the governance programme21 h
  2. Week 2Domain 2: the EU AI Act, its roles and risk categories, the NIST framework and ISO/IEC 4200121 h
  3. Week 3Domain 3: use case, data, training, testing, documentation and release. First module tests21 h
  4. Week 4Domain 4: deployment, vendors, monitoring, incidents and deactivation. Domain tests21 h
  5. Week 5Question structures, the trial exam and its analysis, work on the weak modules, then book the exam21 h

A guideline. The course builds your own plan from your assessment and the hours you have.

Resources

Resources, and what each one is worth

ResourcePriceWorth it?
AIGP body of knowledgeFreeEssential. Work through it competency by competency.
The EU AI Act itselfFreeNo substitute. Read the articles against the recitals.
NIST AI Risk Management FrameworkFreeExaminable, and short enough to read in full.
ISO/IEC 42001Paid standardKnow its purpose and structure. You do not need to buy it for the exam.
IAPP official trainingFrom $995A good introduction. It does not replace practice under time pressure.
IAPP practice examIAPP storeOne exam, useful for the house style of the questions.
22Academy AIGP course€37 a week or €117 a monthThe full course, the exam skills, the measurement and a trial exam. Stop when you are ready.
22Academy trial exam€42, or free with a monthly planFull length under real conditions, with analysis by domain and by question type.
Ten mistakes

Ten mistakes that cost candidates the exam

  1. Treating it as a technical examIt tests governance. You need to know what a model does, not how to build one.
  2. Studying the law onlyThree of the four domains are about running governance in practice.
  3. Mixing up provider and deployerThe obligations follow the role. Settle the role before you read the options.
  4. Skipping the risk categoryWhat an organisation must do depends on the category. Classify first.
  5. Reading the scenario firstThe scenario is written with noise in it. Read the question first.
  6. Missing the qualifierMost, best, primary and except change the logic of the whole question.
  7. Treating the frameworks as oneThe EU AI Act is law, the NIST framework is voluntary and ISO/IEC 42001 can be certified.
  8. Practice in the last week onlyTesting early is what finds the weak domains.
  9. Exam dumpsOften wrong, against the candidate agreement, and useless on scenarios.
  10. Sitting too earlyA retake costs $625. A trial exam costs €42 and tells you whether you are ready.

If you fail, fail cheap. Ten trial exams cost less than one retake.

Buy a trial exam
Jobs and pay

What jobs does AIGP lead to, and what do they pay?

The AIGP is the certification for anyone whose job depends on AI being used responsibly. It leads into AI governance roles, the newest careers in this field, and it adds to the roles that already carry responsibility for data: the data protection officer and the head of privacy are the people organisations turn to first when AI arrives.

Employers ask for it because it shows that you can turn a regulation and a framework into decisions about a real system. Pay rises with the scope of the role.

Each role page shows which certifications it takes, in which order, and how long it takes.

Which first

AIGP or another certification: which first?

CertificationWhat it coversTake it when
AIGPAI governanceYour organisation builds or uses AI
CIPP/EEuropean data protection lawYou advise on the law, or want the legal foundation first
CIPP/USUnited States privacy law, federal and stateYou work with American data, clients or colleagues
CIPMRunning a privacy programmeYou run privacy rather than advise on it
CIPTPrivacy in technologyYou build or run the systems

Not sure? The career plan starts from the job you want.

Glossary

The terms the exam expects you to use precisely

AI system
A machine-based system that infers from its input how to generate outputs such as predictions, content, recommendations or decisions.
Provider
Develops an AI system, or has it developed, and places it on the market or puts it into service under its own name.
Deployer
Uses an AI system under its own authority in a professional setting.
General-purpose AI model
A model that can perform a wide range of tasks and can be built into many downstream systems.
High-risk AI system
A system in a category the EU AI Act lists, or a safety component of a regulated product. The strictest requirements apply.
Prohibited practice
A use of AI that the EU AI Act bans outright.
Conformity assessment
The process that shows a high-risk system meets the requirements before it goes on the market.
Fundamental rights impact assessment
An assessment that certain deployers carry out before first using a high-risk system.
Human oversight
Measures that let people understand, monitor and, where needed, override or stop a system.
Post-market monitoring
The provider's system for collecting and reviewing how a system performs once it is in use.
Serious incident
A malfunction or failure with grave consequences, which must be reported.
AI management system
The policies, roles and processes an organisation uses to govern AI, as set out in ISO/IEC 42001.
Model card
A short document on a model's purpose, its performance and its limits.
Bias
Systematic error that produces unfair outcomes for a group of people.
Explainability
The ability to give the reasons behind an output in terms a person can follow.
Data drift
A change in the input data over time that lowers a model's performance.
Red teaming
Structured adversarial testing to find failures before others do.
Hallucination
An output of a generative model that is fluent and wrong.

Know the terms. Now prove it.

Free account, fifteen minutes, your score per domain.

Exam day

Exam day, and after

Pace

About 1 minute 40 seconds a question. Flag and move on.

Eliminate first

Removing what is wrong is faster than hunting for what is right.

The qualifiers

Most, best, primary and except decide what is being asked.

The break

Fifteen minutes at the midpoint. The first half is submitted for good.

Online or at a test centre: both work. If you sit online, do a full dry run the day before, and disconnect second screens and docking stations.

Passed? You keep the certification with 20 CPE credits in every two-year period, plus the maintenance fee or an IAPP membership that includes it.

Not this time? A retake costs $625, or $475 for IAPP members, and your progress in the course is still there.

“A must have tool for AIGP test preparation”

MG, AIGP

Harder than the real thing, on purpose.

We believe a trial exam should hold a higher standard than the real one. Pass ours at the first attempt, sit the real exam within two weeks and fail it, and we refund what you paid for that trial exam.

Conditions
Questions

Questions people ask about the AIGP

Is the AIGP hard?

It is demanding rather than hard. The material is new to almost everyone: lawyers meet the technology for the first time, and technologists meet the law. Most of the effort goes into vocabulary and roles, knowing exactly what a provider is, which risk category applies and which framework does what. The rest is applying those rules to a messy scenario under a clock, so candidates who practise on exam-style questions early do far better than those who only read.

Do I need a technical background?

No. There are no prerequisites, no degree and no minimum experience. The exam asks what an AI system does and where it can go wrong, not how to build one. Candidates come from privacy, law, compliance, risk, security, product and management, and the course explains every concept in plain language as well as in its formal form.

Do I need the CIPP/E first?

No. The AIGP stands on its own. The CIPP/E does help, because the exam returns to data protection in several places, such as automated decisions and impact assessments. If you work with European data, the two together are a strong combination, in either order.

How is 22Academy different from IAPP training?

The IAPP offers traditional classroom courses and e-learning in which the content is presented to you. We start from the fact that you already have access to all the content, so we focus on getting it into your head rather than throwing more of it at you. We test your knowledge constantly, because that is what makes it stick, and we explain every concept in the form and length that suits you. We also teach the exam skills: how the questions are built, what they are really asking and where the traps are. And we embrace AI for your study, because you use it already; we show you how to use it more effectively.

Do I need a trial exam if I have done practice questions?

Practice questions check your knowledge one item at a time. The trial exam checks whether you can answer one hundred in a row under the clock, the way the real exam asks it. Its analysis shows whether you lose marks on the content or on the type of question, so you know exactly what to work on in the last days. It is the last check before you book.

Can I use my own AI app to prepare?

Yes, and we encourage it. An AI app is a good partner for rehearsing: it will quiz you, rephrase a concept or test you on a passage you give it. It is a poor judge of whether you are ready, because it cannot tell you what you failed to ask. The course gives you prompts that make your own app quiz you properly, and measures you with its own questions.

Is there a free option?

Yes. You can start your course for free and see whether the system suits you: the assessment and the first lesson are open without a card. When you want the rest, the course is €37 a week, or €117 a month with a trial exam included.

Can I stop at any time?

Yes. You pay for the weeks or months you study and cancel whenever you like; your access runs to the end of the period you paid for. If you come back within a year, your progress is still there.

Is 22Academy affiliated with the IAPP?

No. The IAPP owns the AIGP and sets the exam. 22Academy prepares candidates for it independently, and has done so since 2018.

More in the FAQ.

Bas Hennis, founder of 22Academy

Questions about your study?

How to plan around a job, when to book the exam, whether you are ready: connect and ask. Support is free!

info@22academy.com Connect

Start where you actually are

Free account, fifteen minutes, your score per domain.