Skip to content
SimpleAct Logo

Guide · Data protection

As of 6 October 2026 · after the Digital Omnibus

EU AI Act for Data Protection Officers: Tasks, DPIA and the AI Act

Wherever AI processes personal data, the AI Act and the GDPR meet – and the matter almost always lands with data protection. This guide shows what the AI Act changes for DPOs and what it does not, how the DPIA and AI Act duties interlock, where the conflict of interests starts, and what is due in the next 90 days.

  • Art. 37–39 GDPR in an AI context
  • DPIA with Art. 26(9) AI Act
  • Linking FRIA and DPIA
  • Checklist with progress
app.simpleact.de

Record of processing

Linked to the AI inventory

A

2/3

AI systems with RoPA

1

without RoPA entry

1

DPIA approved

1

DPIA with the DPO

AI systems, RoPA and DPIA

  • AI-based application screeningRoPA linked · DPIA residual risk “medium”Approved
  • Service chatbot “Lia”RoPA linked · DPIA awaiting opinionWith DPO
  • Marketing copy generatorNo RoPA entryOpen

Work lists

  • AI systems without RoPA
  • DPIA required?
  • Review dates

Sample data · GDPR module from Professional

In short

The AI Act gives DPOs no new statutory tasks. Their tasks remain those in Art. 39 GDPR: inform, advise, monitor, advise on the DPIA. What is new is the material: deployers of high-risk AI use the provider’s information for their DPIA (Art. 26(9) AI Act), and a fundamental rights impact assessment may refer to the DPIA (Art. 27(4)).

Basics

What the AI Act changes for DPOs – and what it doesn’t

The AI Act (Regulation (EU) 2024/1689) regulates AI systems whether or not they process personal data. The GDPR regulates every processing of personal data, including by AI. Both apply side by side; the AI Act expressly leaves the GDPR unaffected. For DPOs that means: the legal basis of their role remains the GDPR, but the subject matter of their advice gets broader.

The AI Act has no “AI officer” and assigns no tasks of its own to the DPO. That matters, because many organisations quietly assume data protection will “handle the AI thing”. The DPO advises and monitors – they are not responsible for classifying AI systems, approving them or organising AI literacy training under Art. 4 AI Act. Those tasks sit with the organisation as deployer or provider.

Even so, there is hardly a way around data protection. Almost every AI system in an organisation processes personal data – inputs, logs, training data, outputs about people. So it needs a record of processing activities (Art. 30 GDPR), a legal basis, usually a data processing agreement with external vendors (Art. 28) and, where high risk is likely, a data protection impact assessment (Art. 35). For all of this the DPO advises and monitors under Art. 39.

The closest link arises with high-risk AI. Deployers use the information the provider supplies under Art. 13 AI Act to comply with their DPIA duty (Art. 26(9)). And anyone who must carry out a fundamental rights impact assessment under Art. 27 can take over requirements already met by a DPIA by reference (Art. 27(4)). A DPIA set up properly saves duplicate work later.

The Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) moved the high-risk deadlines: Annex III applies from 2 December 2027. GDPR duties, by contrast, have applied unchanged since 2018 – including to every AI system that already processes personal data today.

GDPR: DPO tasks – unchanged, but broader material
Art. 39
GDPR: DPO tasks – unchanged, but broader material
AI Act: DPIA using the provider’s information
Art. 26(9)
AI Act: DPIA using the provider’s information
AI Act: FRIA may refer to the DPIA
Art. 27(4)
AI Act: FRIA may refer to the DPIA

Duties

What the EU AI Act means for data protection officers

The table connects GDPR duties with the points in the AI Act where they meet. The addressee is the organisation in each case; the right-hand column describes the DPO’s role under Art. 39 GDPR.

Scroll the table sideways →

What the EU AI Act means for data protection officers
DutyBasisAddresseeApplies since / fromDPO’s role
Designate a DPO, publish and notify contactArt. 37(1), (7) GDPR · § 38 BDSGControllers, processorssince 25 May 2018In Germany also from 20 people constantly engaged in automated processing, or where processing is subject to a DPIA.
Involve the DPO earlyArt. 38(1) GDPRControllerssince 25 May 2018Be involved in new AI use cases at the request stage, not just before go-live.
Record of processing activitiesArt. 30 GDPRControllerssince 25 May 2018Advise and monitor that every AI system with personal data has a RoPA entry.
Data protection impact assessmentArt. 35 GDPRControllerssince 25 May 2018Give advice (Art. 35(2)) and monitor its performance (Art. 39(1)(c)).
DPIA with provider informationArt. 26(9) AI ActDeployers of high-risk AI2 Dec 2027Make sure the instructions for use and provider information (Art. 13) feed into the DPIA.
Fundamental rights impact assessmentArt. 27 AI Act, para. 4Public bodies, private providers of public services, Annex III 5(b)/(c)2 Dec 2027Advise which parts of the DPIA can be taken over by reference; the FRIA itself is the deployer’s duty.
Automated individual decisionsArt. 22 GDPRControllerssince 25 May 2018Have it checked whether an AI output is in effect a solely automated decision with significant effect.
Informing data subjectsArt. 13, 14 GDPR · Art. 26(11) AI ActControllers / deployersGDPR since 2018, AI Act from 2 Dec 2027Align privacy notices with the AI-specific information for Annex III decisions.
Explanation of decisionsArt. 86 AI ActDeployers (Annex III)2 Dec 2027Design the process for requests together with the access process under Art. 15 GDPR.
Emotion recognition, biometric categorisationArt. 50(3) AI ActDeployers2 Aug 2026Monitor that data subjects are informed and processing complies with the GDPR.
Keep logs – and limit themArt. 26(6) AI Act · Art. 5(1)(e) GDPRDeployers of high-risk AI2 Dec 2027Advise how the six-month minimum fits with storage limitation.
Processing agreements with AI vendorsArt. 28 GDPRControllerssince 25 May 2018Advise on the DPA review: sub-processors, third countries, instructions, the vendor’s own use of data.

Simplified overview. The duties under Art. 26, 27 and 86 AI Act apply only to high-risk AI under Annex III from 2 Dec 2027. The GDPR already applies today to every AI system with personal data.

In detail

Six topics for data protection officers

Each topic describes the legal position, what it means for the DPO’s work and which recommendations have proven themselves.

01

The tasks under Art. 39 GDPR in an AI context

Art. 39(1)(a)–(e), (2) GDPR

Art. 39 GDPR lists the minimum tasks: inform and advise the controller and employees of their obligations (a), monitor compliance with the GDPR and data protection policies, including assignment of responsibilities, awareness-raising and training (b), advise on the DPIA where requested and monitor its performance (c), cooperate with the supervisory authority (d) and act as its contact point (e). Under para. 2 the DPO has due regard to the risk associated with processing operations.

For AI that means two things. First: the risk-based approach permits and demands priorities. A text generator in marketing needs less attention than a system that pre-sorts job applications. Second: “monitoring” refers to data protection, not to the entire AI Act. The DPO does not have to check whether a chatbot meets the Art. 50 notice duty – but can raise it if they notice.

  • Inform and advise (a)
  • Monitor, including training (b)
  • Advise on and monitor the DPIA (c)
  • Prioritise by risk (para. 2)
02

Conflict of interests: advise yes, decide no

Art. 38(3), (6) GDPR

The DPO may fulfil other tasks and duties as long as they do not result in a conflict of interests (Art. 38(6)). They receive no instructions on the exercise of their tasks, may not be dismissed or penalised for performing them, and report directly to the highest management level (Art. 38(3)).

With AI the conflict is close at hand. A DPO who also decides on approving an AI tool would later have to monitor their own decision. The same applies where the DPO determines the purposes and means of processing, for example as project lead for introducing an AI system. Advice, by contrast, is unproblematic – early, in writing, with a clear recommendation. Whether the DPO can take on additional operational tasks in AI governance depends on whether such decisions are part of them. Recommendation: in the RACI matrix the DPO appears only as C or I on decisions.

  • No approvals by the DPO
  • No project lead for the AI rollout
  • Early, documented advice instead of co-decision
03

Linking the DPIA and the AI Act

Art. 35 GDPR · Art. 26(9), Art. 13 AI Act

A DPIA is required where processing is likely to result in a high risk to the rights and freedoms of natural persons (Art. 35(1)). Expressly named are, among others, a systematic and extensive evaluation of personal aspects based on automated processing (para. 3(a)) and large-scale processing of special categories (para. 3(b)). Many high-risk use cases in Annex III – candidate selection, performance assessment, creditworthiness – typically meet such criteria.

The AI Act adds to the evidence base: deployers of high-risk AI use the information the provider supplies under Art. 13 – instructions for use, performance limits, guidance on input data and human oversight – to comply with their DPIA duty (Art. 26(9)). For the DPO that is a concrete lever: if this information is missing, the DPIA is incomplete, and that belongs in the opinion.

The minimum content remains the GDPR’s (Art. 35(7)): a systematic description of the operations and purposes, an assessment of necessity and proportionality, an assessment of the risks to data subjects and the measures envisaged. If high risk remains after the measures, the supervisory authority must be consulted beforehand (Art. 36). And the DPIA is not a one-off document: where the risk changes – for example through a new model – it must be reviewed (Art. 35(11)).

  • Request provider information under Art. 13 AI Act
  • Minimum content under Art. 35(7) GDPR
  • Review on model change (Art. 35(11))
04

Fundamental rights impact assessment: reference instead of duplication

Art. 27(1)–(4) AI Act

The fundamental rights impact assessment (FRIA) is a duty of the AI Act, not of data protection law. Before first putting a high-risk system into use it applies to bodies governed by public law and private entities providing public services, and to deployers of systems under Annex III point 5(b) and (c) (creditworthiness, risk assessment and pricing in life and health insurance) – except systems under Annex III point 2. The results must be notified to the market surveillance authority.

In substance the FRIA overlaps with the DPIA but is broader: it looks at fundamental rights as a whole, even without personal data, and requires details such as the period and frequency of use, affected groups, human oversight and complaint mechanisms. Art. 27(4) allows requirements already met by a DPIA to be taken over by reference. Recommendation: for affected systems, build the DPIA from the start so the FRIA structure can be found in it.

  • Only certain deployers are obliged
  • The FRIA is broader than the DPIA
  • Reference to the DPIA expressly allowed (para. 4)
05

Automated decisions and the right to an explanation

Art. 22 GDPR · Art. 26(11), Art. 86 AI Act

Art. 22 GDPR gives data subjects the right not to be subject to a decision based solely on automated processing which produces legal or similarly significant effects – with exceptions. With AI, practice decides: a “human review” that merely rubber-stamps outputs does little to change the automated character. This is where the GDPR meets human oversight under Art. 14 and Art. 26(2) AI Act.

The AI Act adds two rights for Annex III systems: deployers inform people about whom a high-risk system makes or assists decisions (Art. 26(11)), and affected persons can request an explanation of the role of the AI system and the main elements of the decision where it has legal or similarly significant effects (Art. 86). Recommendation: run these requests through the same channel as access requests under Art. 15 GDPR.

  • Rubber-stamping is not a human decision
  • Information under Art. 26(11) AI Act
  • Bundle the Art. 86 explanation right with Art. 15 GDPR
06

RoPA, DPA and logs: the day-to-day

Art. 28, 30, 5(1)(e) GDPR · Art. 26(6) AI Act

Most of the work is unspectacular. Every AI system with personal data needs a RoPA entry with purpose, legal basis, data categories, recipients and retention period. A field for automated decisions is particularly valuable: where it says “yes”, the system usually has heightened requirements. Ideally each RoPA entry is linked to the matching AI inventory record – then you see the processing for every system and the system for every processing.

With external vendors the DPO advises on the DPA review: sub-processors, third-country transfers, instructions and above all whether the vendor uses inputs for its own purposes such as model training. On logs: the AI Act requires deployers of high-risk AI to keep them for at least six months, the GDPR requires storage limitation. Both can be reconciled – with a fixed period per system that covers the minimum, and clear access roles.

  • A RoPA entry per AI system, linked to the inventory
  • DPA: clarify the vendor’s use of data
  • Log period: cover the minimum, do not exceed it

RACI excerpt

Where the DPO is involved – and where they shouldn’t decide

Excerpt from an AI governance RACI matrix with the tasks data protection takes part in. The highlighted column shows the role of the data protection officer. Each task has exactly one A. The assignment is a suggestion.

Scroll the table sideways →

Where the DPO is involved – and where they shouldn’t decide
TaskManagementAI governance leadBusiness unit / system ownerIT lead / CTOData protection (DPO)Legal
Create the RoPA entry for an AI systemArt. 30 GDPRICA/RCC–
Check whether a DPIA is neededArt. 35(1), (3) GDPRICA/RCCC
Carry out the DPIA incl. provider informationArt. 35 GDPR · Art. 26(9) AI ActACRCCC
Give an opinion on the DPIAArt. 35(2), Art. 39(1)(c) GDPRIII–A/R–
Prepare the FRIA, refer to the DPIAArt. 27(4) AI ActARR–CC
Review the DPA with the AI vendorArt. 28 GDPRICRCCA
Approval decision (go/no-go)Internal processARCCCC
Monitor compliance, check trainingArt. 39(1)(b) GDPRICC–A/R–
Assess and notify a data breachArt. 33 GDPRARRRCC
  • RResponsible – does the work
  • AAccountable – owns and decides
  • CConsulted – asked beforehand
  • IInformed – kept informed
  • A/R for the DPO appears only for their own tasks under Art. 39 GDPR – opinion and monitoring. On decisions about processing they are C, so that no conflict of interests arises under Art. 38(6) GDPR.
  • The DPIA is the controller’s duty. The DPO advises and monitors but does not write it – even though that is often expected in practice.
  • The full matrix with twelve tasks including the works council is in the AI governance framework.

Interactive

Top tasks for DPOs in the next 90 days

Twelve tasks in three phases. Tick off what is done – progress shows at the top. Expand any task for details.

0 of 12 tasks done

0%

Your progress is stored in this browser only, never transmitted.

Days 1–30

0/4

Overview

  • Reconcile the AI inventory with the RoPA
  • Clarify your own role in AI governance
  • Anchor early involvement in the request process
  • Flag potential high-risk systems

Days 31–60

0/4

DPIA and contracts

  • Record the DPIA need per AI system with reasons
  • Have provider information under Art. 13 AI Act requested
  • Have DPAs with AI vendors reviewed
  • Make the DPIA structure FRIA-ready

Days 61–90

0/4

Operations and training

  • Align log periods with the retention policy
  • Extend privacy notices for AI use
  • Bring data protection content into AI training
  • Set review dates for DPIA and RoPA

Order and time frame are a recommendation, not a legal requirement. External DPOs adapt the tasks to their mandate.

From the DPO’s point of view

Questions DPOs get asked

Typical questions from business units, management and the works council – with answers you can use as a starting point.

“So you’re in charge of AI now too?”

For data protection in AI yes, for the AI Act as a whole no. The AI Act assigns no tasks to the DPO. Classification, approval and organising training sit with the organisation; the DPO advises and monitors under Art. 39 GDPR.

AI governance framework

“Do we need a DPIA for ChatGPT?”

Not automatically. It depends on purpose and data. For texts without personal data usually not; for evaluating applications or large volumes of customer data very likely. The decision belongs on file with reasons – even if it is “no”.

DPIA guide

“Is the DPIA enough, or do we also need a FRIA?”

Only certain deployers of high-risk AI need a FRIA under Art. 27 AI Act, such as public bodies or banks assessing creditworthiness. Those who need one can take over parts of the DPIA by reference (Art. 27(4)).

“May the vendor use our inputs for training?”

Only if that is settled cleanly in the contract and under data protection law. Many standard contracts let you opt out. Without clarity the vendor may no longer be acting only as a processor – that belongs in the DPA review.

Data processing agreements

“Can’t you just approve the tool?”

Better not. Whoever approves would later have to monitor their own decision – a conflict of interests under Art. 38(6) GDPR. The DPO gives a written recommendation; management or a delegated function decides.

“How long may we keep the AI logs?”

As long as necessary, as short as possible. For high-risk AI the AI Act requires at least six months for logs under your control. A fixed period per system in the retention policy creates clarity.

Retention policy

Deadlines

Dates from the AI Act and the GDPR

The GDPR has applied since 2018 to every AI system with personal data. The AI Act adds dates that touch data protection.

  1. 25 May 2018

    GDPR applies

    RoPA, legal basis, DPA, DPIA – also for every AI system with personal data.

  2. 2 Feb 2025

    Prohibited practices (Art. 5) and AI literacy (Art. 4)

    Bring data protection content into AI training; prohibited practices often involve biometric data.

  3. 2 Aug 2026

    Transparency duties under Art. 50

    Align information on emotion recognition and biometric categorisation (para. 3) with GDPR notices.

  4. 2 Dec 2026

    New prohibitions (Omnibus)

    Non-consensual intimate deepfakes of identifiable people – a data protection issue of the first order.

  5. 2 Dec 2027

    High-risk duties for Annex III systems

    DPIA with provider information (Art. 26(9)), FRIA (Art. 27), information (Art. 26(11)), explanation (Art. 86).

  6. 2 Aug 2028

    High-risk duties for Annex I systems

    AI in regulated products, such as medical devices with health data.

Status after the Digital Omnibus, Regulation (EU) 2026/1744, in force since 27 July 2026. All AI Act deadlines →

With SimpleAct

Data protection and the AI inventory in one system

In SimpleAct’s GDPR module every RoPA entry and every DPIA can be linked to an AI system. The view on the right shows the audit log with sample data – every change, including an uploaded DPIA report, is recorded.

  • Record of processing with AI link

    Professional

    RoPA entries with purpose, legal basis (the six values of Art. 6(1)), data categories, automated decisions and linked AI system. A work list shows AI systems without a RoPA entry.

  • DPIA as its own record

    Professional

    Necessity check with reasons, risks with likelihood and severity, measures, residual risk, DPO opinion and approval. A work list shows high-risk systems and systems with personal data without a linked DPIA.

  • TOMs, DPAs, data subject requests, breaches, retention

    Professional

    The other parts of the GDPR module in the same system as the AI inventory.

  • DPO management

    Professional

    DPO profile with appointment and notification date (Art. 37(7)) and task allocation under Art. 39.

  • Third-country transfers, authority packages

    Enterprise

    Transfer register with review status and TIA wizard (Art. 44 to 49) and response packs for supervisory authority requests.

  • Audit trail

    all plans

    Every change with person, time, old and new value; built as a hash chain with an integrity check.

Plan assignment per the pricing page. Application and database stay in Germany (Hetzner, Nuremberg; backups in Falkenstein). Sub-processors are Hetzner, Cloudflare (file storage), Stripe (payments), Brevo (email) and Anthropic (AI features only after opt-in; support chat see privacy policy); details in the DPA.

app.simpleact.de

Audit Log & Export

247 of 312 entries

A

Audit readiness

Audit-ready

Complete

Partial

Missing

EU Declaration of Conformity (Art. 47)

PDF · DE/EN · incl. audit history

ActionEntityEvent
UPDATERisk AssessmentRe-assessment completed
CREATEEvidenceEvidence "DPIA report" added
UPDATEGovernanceApproved by reviewer
CREATEAI SystemSystem "Chatbot Lia" created

FAQ

Frequently asked questions from data protection officers

Not as such. The AI Act assigns no tasks of its own to the DPO and does not require an AI officer. The DPO’s tasks still follow from Art. 39 GDPR: inform, advise, monitor and advise on the DPIA. Because almost every AI system processes personal data, the DPO is nonetheless involved in most AI projects.
That depends on whether the role includes decisions on the purposes and means of processing or on approvals. The DPO may fulfil other tasks as long as they do not result in a conflict of interests (Art. 38(6) GDPR). Whoever approves AI tools would later have to monitor their own decision. A separation is advisable: the DPO advises, decisions are taken elsewhere.
No. A DPIA is required where processing is likely to result in a high risk to the rights and freedoms of natural persons (Art. 35(1) GDPR). For many high-risk uses under Annex III AI Act, such as recruiting or credit checks, that is typically the case. Either way the decision should be documented with reasons.
Deployers of high-risk AI use the information the provider supplies under Art. 13 AI Act to comply with their DPIA duty. In practice: instructions for use, performance limits and guidance on input data and oversight belong in the DPIA. If they are missing, the DPO should say so in the opinion.
The FRIA under Art. 27 AI Act is a separate duty of certain deployers of high-risk AI and looks at fundamental rights as a whole. It overlaps with the DPIA. Art. 27(4) expressly allows requirements already met by a DPIA to be taken over into the FRIA by reference. One cannot replace the other.
Under Art. 37(1) GDPR among others for public authorities, for large-scale regular and systematic monitoring or large-scale processing of special categories. In Germany also under § 38 BDSG where as a rule at least 20 people are constantly engaged in automated processing, or regardless of that where processing is subject to a DPIA. Contact details must be published and notified to the supervisory authority (Art. 37(7)).
No. Art. 4 AI Act obliges providers and deployers to take AI literacy measures. Under Art. 39(1)(b) GDPR the DPO monitors data protection awareness and training and can contribute content on handling personal data in AI tools. Organising AI training is the organisation’s job.
Deployers of high-risk AI must keep automatically generated logs under their control for at least six months (Art. 26(6) AI Act). The GDPR requires data not to be kept longer than necessary (Art. 5(1)(e)). A fixed period per system that covers the minimum, plus tight access rights, reconciles both.
Besides Art. 22 GDPR (no solely automated decision with significant effect, with exceptions) and the right of access under Art. 15 GDPR, the AI Act brings for Annex III systems an information duty of the deployer (Art. 26(11)) and a right to an explanation of the role of the AI system and the main elements of the decision (Art. 86).
Yes, in the GDPR module from the Professional plan. The DPIA is its own record with necessity check, risks, measures, residual risk, opinion and approval, and can be linked to the RoPA entry and AI system. A work list shows systems without a linked DPIA. It does not replace your own assessment.
No. The AI Act does not replace the data protection officer and does not prescribe an “AI officer”. The duty to designate a DPO continues to follow from Art. 37 GDPR and § 38 BDSG. Interfaces exist in the DPIA (Art. 26(9) AI Act) and the fundamental rights impact assessment (Art. 27(4) AI Act), for Annex III high-risk AI from 2 December 2027 in each case; Art. 26(9) applies to Annex I high-risk AI from 2 August 2028.
Ask the vendor about its sub-processors with location and role, about training on your data, logs and third-country transfers. The vendor may engage further processors only with your prior specific or general written authorisation; under a general authorisation it must inform you of changes so you can object (Art. 28(2) GDPR). It must impose the same data protection obligations on them by contract (Art. 28(4)). Our page on data processing includes a checklist for AI vendors; SimpleAct manages DPAs in the GDPR module from Professional.

More questions? We're happy to help. Send email · Try 7 days (credit card required)

Sources and status

Last reviewed on · SimpleAct editorial team

This page is for orientation and is not legal advice. The text published in the Official Journal of the EU is authoritative. RACI assignment, checklist and time frames are recommendations, not legal requirements.

Changes to this page

  • 30 October 2026 – New answers to frequently asked questions added.
  • 6 October 2026 – Rebuilt as a guide: tasks under Art. 37–39 GDPR, conflict of interests (Art. 38(6)), DPIA linked with Art. 26(9) and Art. 27(4) AI Act, RACI excerpt, interactive 90-day checklist, ten FAQs. Hosting statement corrected (Germany only).
  • 18 May 2026 – First version as a role page for data protection officers.

RoPA, DPIA and AI inventory linked

SimpleAct connects the GDPR module with the AI inventory – so you see the processing for every AI system and the system for every processing.