GUIDE · AI ACT

AI Act and GDPR: differences, overlaps and how to coordinate them

After this guide you will be able to distinguish what the GDPR regulates from what the AI Act regulates, identify where processing of personal data overlaps, use criteria to compare the two sets of obligations, and organise your work so that a single analysis serves both. The comparison is general: specific cases should be verified with a consultant.

Updated october 2026
7 min read
Altovar Technical Editorial Team

1. Two regimes with different subjects

The GDPR, the regulation on the protection of personal data, governs the processing of personal data regardless of the technology used. A customer database, a spreadsheet or an AI system that processes the same information is subject to the same basic rules: a lawful basis, clear purposes, data minimisation, security and respect for individuals' rights. This distinction should be clarified before any other analysis.

The AI Act, by contrast, regulates artificial intelligence systems according to risk and use. It is not limited to personal data: a system can be subject to obligations even when it works on non-personal data, and a processing of personal data may fall outside the AI Act if no AI system is involved. The point of intersection is when an AI system processes personal data, which is the most frequent case in companies.

2. Where they overlap

The most common overlaps concern transparency towards individuals, risk assessment, technical and organisational documentation, management of training data and human oversight of decisions. In many cases the same document can serve both regimes, if it is prepared with care and updated together. It depends on the system and its context, so each case has to be read on its own merits.

The overlap also creates tensions. The GDPR's data minimisation principle pushes towards reducing data, while the data quality required by an AI system may push towards collecting more. This tension should be resolved in writing, with a reasoned choice, and not left to the technician configuring the tool. The decision should be justified in writing and reviewed when the tool changes.

3. Comparison on the main points

The table compares the two regimes on the points companies most often encounter. The answers are generic and must be adapted to the specific case with a consultant. It is not an exhaustive list: it is there to orient you, not to close the work. Read it as a map to be completed with the official text and with the help of an expert. The rows are written to be compared, not to be copied into an internal document without verification.

The comparison helps you decide which document serves which question, and avoids asking the same text for answers it cannot give. In practice, a GDPR document is not enough to answer questions about AI, and vice versa. When a topic touches both, write a line linking the two documents, so that nobody stops halfway.

AspectGDPRAI Act
Subject matterProcessing of personal dataAI systems, based on risk and use
Logic of the rulePrinciples of lawfulness, minimisation and securityRisk classification, prohibitions and graduated obligations
Typical documentationRecord of processing activities and data protection impact assessmentTechnical documentation, classification register and, in some cases, fundamental rights impact assessment
People involvedData subjects whose data is processedPeople affected by the outputs and workers exposed to the system
Supervisory authorityData protection authorityDesignated AI supervisory authorities

4. A shared check for sensitive systems

Set up a single check for systems that process personal data: description of the system, purpose, data, individuals involved, suppliers and main risks. This sheet serves both regimes. The data protection officer and the person responsible for the AI system read it together, each from their own perspective, and the sheet becomes the shared point of reference. A name and a role for each entry make the sheet usable.

Then separate what is specific. For the GDPR you need the lawful basis and the information to individuals. For the AI Act you need the risk classification and oversight. Not everything should be shared. A single document that mixes everything becomes hard to maintain and hard to show to whoever is checking. The starting point is always the purpose stated in the sheet. This avoids having to reconstruct everything when an authority asks for clarification.

5. Common mistakes

The first mistake is treating the AI Act as a technical matter and the GDPR as a legal one, with two owners who do not talk to each other. The result is double documentation with contradictory content. The second is assuming that GDPR compliance is enough to cover the AI Act: they are different rules, with additional obligations that the GDPR does not provide for.

The third is forgetting cases where the system does not process personal data directly, but produces it indirectly, for example when an output can identify a person. The fourth is updating only one register when the tool changes. Every change should be reflected in both documents, with the same date. A simple check is to ask, with every change, which documents need updating.

6. How to decide and what to ask the supplier

When you select or assess an AI supplier, ask for information that both regimes require: where data is processed, who has access to it, whether inputs are used to train the model, how it is protected, what information it provides about how the system works and its limitations, and how it reports incidents. A written answer on these points is worth more than a generic statement of compliance.

To decide whether a processing activity is acceptable, start from its purpose and the risk to people, not from the technology. A system with a clear purpose, limited data and genuine oversight is easier to reconcile than one that does everything. Record the decision with its reasoning, so that it remains explainable over time. Also ask what impact the choice has on data quality and on human oversight.

7. How to start

This week, list the systems that use AI and process personal data. Next to each one, note whether you already have a GDPR analysis and who is responsible for it. This column immediately shows you where duplications or gaps exist, and where an existing document can be reused. You do not need a complete sheet to begin. What you need is to appoint an owner and to know which systems process personal data.

Then arrange a meeting between the data protection officer, whoever manages IT and the compliance consultant. The goal is a single shared sheet for the most sensitive systems. It does not need to be complete straight away: what matters is that the two regimes speak the same language. The first meeting can be short, but it must close with an action for each participant.

Frequently Asked Questions

If I comply with the GDPR, am I covered for the AI Act too?

No. The GDPR remains necessary, but the AI Act adds rules on classification, prohibitions, transparency, oversight and documentation of AI systems, even where no personal data is involved. Complying with the GDPR is a foundation, not complete coverage. Check with your consultant which AI Act obligations apply to you.

Who should manage compliance with both regimes?

In many organisations the data protection officer oversees the GDPR, while the AI Act requires technical and organisational expertise on how the systems are used. It need not be the same person, but you do need a coordination point with a clear mandate, otherwise decisions remain stuck between two functions.

Can I use a GDPR data protection impact assessment for the AI Act?

In part. The data protection impact assessment can be the basis, but the one required by the AI Act has, in some cases, different content, such as the broader impact on fundamental rights and the oversight of the system. Complete the existing document with the missing points rather than duplicating it.

Is a system that uses anonymous data excluded from the GDPR?

It depends on how truly anonymous the data is. Genuinely anonymous data falls outside the GDPR, but in practice it is often only pseudonymised or can be combined with other sources. Do not take anonymisation for granted: verify it with your consultant. In any case the AI Act still applies if the system falls within its scope.

Where should I start if I already use many tools?

Start with the inventory and with the systems that affect people the most. For each one, check whether a GDPR analysis exists and whether the system falls into a sensitive use under the AI Act. Work first on cases involving decisions about people, then on the others. Do not try to close everything at once, because the result would be a superficial document.

About this content

General informational guide. Regulations and technical requirements change over time: always check the official sources and assess your specific case with a qualified technical or legal advisor.

GET IN TOUCH

Talk to our team.

We'll show how Altovar integrates into your infrastructure. No commitment, no sales pitch. Just a technical conversation.

BOOK A DEMO →