# GDPR AI Compliance: What the GDPR Requires for AI Training, Deployment and Automated Decisions, Plus Switzerland

Source: https://superintelligencecompliance.ai/guides/gdpr-ai-compliance  
Publisher: ZSKFL Management  
Last updated: 2026-11-12. Not legal advice.

> GDPR duties for AI by article, how EDPB Opinion 28/2024 reads them, Article 22 after SCHUFA, fines, AI Act overlap, the unadopted Omnibus, Swiss FADP.

## At a glance

- The GDPR has no AI chapter, and as of November 12, 2026 its text had not been amended. Its existing rules govern AI processing of personal data within its scope.
- EDPB Opinion 28/2024 (adopted December 17, 2024) says AI models trained with personal data cannot, in all cases, be considered anonymous, recalls the three-step legitimate interest test, and says authorities should take into account whether a company deploying another company’s model conducted an appropriate assessment of whether it was developed by unlawfully processing personal data. The opinion is the Board’s reading, not GDPR text.
- The Court of Justice has held that a credit information agency’s automated credit score is an Article 22(1) decision where a third party to which it is transmitted draws strongly on it to establish, implement or terminate a contract with that person (*SCHUFA*, 2023), and that, for automated decisions under Article 22(1), Article 15(1)(h) lets the person require the controller to explain the procedure and principles actually applied to obtain a specific result (*Dun & Bradstreet Austria*, 2025).
- The Commission’s Digital Omnibus proposal of November 19, 2025, COM(2025) 837, would, if adopted, add AI-specific GDPR rules, including a new Article 88c on legitimate interest for AI. The Council’s negotiating mandate of October 21, 2026 keeps the current definition of personal data; as of November 12, 2026, Parliament had not adopted its position. It is not law.

## What does the GDPR require when a business uses AI?

Within its scope, the GDPR applies whenever personal data is processed to build, train, test or run an AI system, and, in the EDPB’s view, to a model trained on personal data unless the model is anonymous. A business needs a lawful basis for each purpose, must meet the Article 5 principles, must tell people how their data is used unless an exception applies, must honour access, erasure and objection rights, and may take solely automated decisions with legal or similarly significant effects only under the Article 22 exceptions and safeguards.

The GDPR also reaches controllers and processors not established in the Union when their processing of data about people in the Union relates to offering them goods or services or to monitoring their behaviour as far as it takes place in the Union (Article 3(2)). Fines reach €20 million or 4% of worldwide annual turnover, whichever is higher (Article 83).

**GDPR provisions that apply to AI, as of November 12, 2026**

| GDPR provision | What it requires for AI | Key EDPB or court source |
|---|---|---|
| Article 4(1) and Recital 26: personal data | A model trained on personal data is outside the GDPR only if it is anonymous; the EDPB says supervisory authorities should take the controller’s documentation into account when assessing that claim (paragraphs 56 and 57) | EDPB Opinion 28/2024; EDPB Guidelines 02/2026 on Anonymisation (version 1.0, for public consultation) |
| Article 6: lawful basis | A lawful basis for each purpose, in development and in deployment; legitimate interest needs the three-step test. Reusing data for a new purpose, such as training, that does not rest on consent or on a Union or Member State law that is a necessary and proportionate measure in a democratic society to safeguard the objectives in Article 23(1) needs a compatibility test that takes into account, among other things, the link between the purposes, the context of collection, the nature of the data, the possible consequences and appropriate safeguards (Article 6(4)) | EDPB Opinion 28/2024 |
| Article 5(1)(b), (c), (d): purpose limitation, data minimisation, accuracy | No further processing in a manner incompatible with the original purposes; data adequate, relevant and limited to what is necessary; data accurate and, where necessary, kept up to date, with every reasonable step taken to erase or rectify inaccurate data without delay, having regard to the purposes | EDPB Opinion 28/2024, paragraph 64 (purpose limitation and minimisation) |
| Article 9: special categories | Processing of data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs or trade union membership, genetic data, biometric data for the purpose of uniquely identifying a person, and data concerning health, sex life or sexual orientation is prohibited unless one of the ten exceptions in Article 9(2) applies | AI Act Article 4a for bias detection |
| Articles 13 and 14: information | Notice of purposes, legal basis, recipients, retention and rights; for data not obtained from the person, also the categories and source, including whether it came from publicly accessible sources; for Article 22 decisions, meaningful information about the logic involved and the significance and envisaged consequences | EDPB Opinion 28/2024, paragraph 63 (Article 14(5)(b) and scraped data) |
| Article 15: access | Confirmation, a copy of the data, and the Article 15(1) information, including for Article 22 decisions meaningful information about the logic involved | CJEU C-203/22, *Dun & Bradstreet Austria* |
| Article 17: erasure | Erasure without undue delay on the listed grounds, among them that the data was unlawfully processed, or that the person objected under Article 21(1) and there are no overriding legitimate grounds | Text of Article 17 |
| Article 21: objection | Where processing rests on Article 6(1)(e) or (f), the controller must stop on an objection unless it demonstrates compelling legitimate grounds that override the person’s interests, rights and freedoms, or the processing is for legal claims; objection to direct marketing, including related profiling, may be made at any time and ends processing for that purpose | Text of Article 21 |
| Article 22: automated decisions | Right not to be subject to a decision based solely on automated processing, including profiling, with legal or similarly significant effects, subject to three exceptions and safeguards | CJEU C-634/21, *SCHUFA* |
| Article 25: data protection by design and by default | Appropriate technical and organisational measures, such as pseudonymisation, both when the means of processing are determined and during processing; by default only the data necessary for each specific purpose | Text of Article 25 |
| Article 28: processors | Use only processors providing sufficient guarantees to implement appropriate technical and organisational measures so that processing meets the GDPR’s requirements and protects data subjects’ rights, under a contract containing the Article 28(3) terms | Text of Article 28 |
| Article 35: DPIA | Assessment before processing, in particular using new technologies, likely to result in a high risk; required in particular for a systematic and extensive evaluation of personal aspects based on automated processing, including profiling, on which decisions with legal or similarly significant effects are based; for large-scale processing of special-category data or data on criminal convictions and offences; and for large-scale systematic monitoring of a publicly accessible area (Article 35(3)) | AI Act Articles 26(9) and 27(4) refer to it |
| Articles 44 to 49: transfers | An adequacy decision, appropriate safeguards with enforceable rights and effective remedies, or an Article 49 derogation for personal data sent outside the EEA | Implementing Decision (EU) 2023/1795; General Court T-553/23, *Latombe* |
| Article 83: fines | Infringing, among other provisions, the principles, lawful basis, special-category rules, data subject rights or transfer rules: up to €20 million or, for an undertaking, 4% of total worldwide annual turnover of the preceding financial year, whichever is higher (Article 83(5)). Infringing, among others, the controller and processor duties in Articles 25 to 39: up to €10 million or 2%, whichever is higher (Article 83(4)) | Text of Article 83 |

## When is an AI model itself personal data?

In Opinion 28/2024, the EDPB said that for a model to be anonymous, both the likelihood of direct, including probabilistic, extraction of personal data about the people whose data was used to develop it and the likelihood of obtaining such data from queries, intentionally or not, should be insignificant, taking into account all the means reasonably likely to be used by the controller or another person.

EDPB Guidelines 02/2026 on Anonymisation, adopted July 7, 2026 as version 1.0 for public consultation (now closed), say anonymity should be assessed from each relevant entity’s perspective. As of November 12, 2026, no later version was listed.

## What lawful basis covers AI training and deployment?

Opinion 28/2024 recalls there is no hierarchy between the Article 6(1) bases. For legitimate interest under Article 6(1)(f), which public authorities cannot use in the performance of their tasks, the Opinion recalls three steps:

- The interest must be lawful, clearly and precisely articulated, and real and present, not speculative.
- The processing must allow the interest to be pursued, with no less intrusive way to pursue it, judged with attention to the amount of personal data and data minimisation.
- The interest must not be overridden by data subjects’ interests or fundamental rights and freedoms, judged in part by their reasonable expectations, including whether the data was publicly available, the context and source of collection, and the possible further uses of the model.
Where data subjects’ interests seem to override the controller’s, the controller may consider mitigating measures; for web scraping, the Opinion’s examples include respecting robots.txt or ai.txt files and keeping an opt-out list.

**If the model was trained unlawfully.** The Opinion gives three scenarios. Where personal data stays in the model and the same controller deploys it, the effect of the unlawful development is assessed case by case. Where another controller deploys it, authorities should take into account whether that deployer conducted an appropriate assessment, as part of its accountability duties under Articles 5(1)(a) and 6, to ascertain that the model was not developed by unlawfully processing personal data. Where it can be demonstrated that the anonymised model’s operation does not involve processing personal data, the GDPR would not apply to that operation, though it applies to personal data later collected in deployment.

## Can special-category data be used to test AI for bias?

Article 9 prohibits processing special-category data unless an exception applies, and the EU AI Act adds one route. The Digital Omnibus on AI, Regulation (EU) 2026/1744, deleted AI Act Article 10(5) and re-enacted its exception as Article 4a. Providers of high-risk AI systems may exceptionally process special-category data to the extent strictly necessary for bias detection and correction under Article 10(2)(f) and (g), subject to appropriate safeguards for the fundamental rights and freedoms of natural persons and, in addition to the GDPR, only if all six conditions are met, among them that other data, including synthetic or anonymised data, would not do. Article 4a(2) extends this to providers and deployers of other AI systems and models and to deployers of high-risk systems, where strictly necessary for bias detection and correction in view of possible biases likely to affect health and safety, have a negative impact on fundamental rights or lead to discrimination prohibited under Union law, and only if all the same conditions and safeguards are applied. It creates no obligation to conduct bias detection and correction. The [AI data governance guide](https://superintelligencecompliance.ai/guides/ai-data-governance) lists the six conditions.

## Can a company rely on the Article 14(5)(b) exception for scraped data?

For data collected from publicly accessible sources, such as by web scraping, the EDPB says reliance on the exception is strictly limited to when its requirements are fully met (Opinion 28/2024, paragraph 63). Article 14(5)(b) lifts the duty to inform people whose data was not obtained from them where doing so proves impossible or would involve a disproportionate effort, in particular for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes subject to Article 89(1), or is likely to render impossible or seriously impair the achievement of the objectives of the processing. In such cases the controller must take appropriate measures to protect the data subject’s rights and freedoms and legitimate interests, including making the information publicly available.

## When is an AI output an Article 22 decision?

Article 22(1) gives a person the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning them or similarly significantly affects them; the Court of Justice reads it as a prohibition in principle (*SCHUFA*, paragraph 52). The right does not apply where the decision is necessary for entering into or performing a contract between the person and the controller, is authorised by Union or Member State law to which the controller is subject and which lays down suitable safeguards, or is based on explicit consent. Under the contract and consent exceptions, the controller must implement suitable measures to safeguard the person’s rights and freedoms and legitimate interests, at least the right to obtain human intervention on the part of the controller, to express a point of view and to contest the decision (Article 22(3)). Such decisions may not be based on special-category data unless Article 9(2)(a) or (g) applies and suitable safeguards are in place (Article 22(4)).

- ***SCHUFA*, Case C-634/21 (December 7, 2023).** A credit information agency’s automated establishment of a probability value about a person’s ability to meet future payment commitments is automated individual decision-making under Article 22(1) where a third party to which it is transmitted draws strongly on it to establish, implement or terminate a contract with that person.
- ***Dun & Bradstreet Austria*, Case C-203/22 (February 27, 2025).** For automated decision-making, including profiling, under Article 22(1), Article 15(1)(h) lets the person require the controller to explain, by means of relevant information and in a concise, transparent, intelligible and easily accessible form, the procedure and principles actually applied to obtain a specific result, such as a credit profile. Neither the mere communication of a complex mathematical formula, such as an algorithm, nor the detailed description of all the steps in automated decision-making satisfies this (paragraph 59). Where the controller says the explanation contains third-party data or trade secrets, it must provide that information to the competent supervisory authority or court, which balances the rights and interests at issue.

## What do the vendor, DPIA and transfer rules add?

An AI vendor that processes personal data on the customer’s behalf is a processor under Article 28; a vendor that determines the purposes and means of processing, such as using customer data for its own purposes, is a controller for that processing (Articles 4(7) and 28(10)). The [AI vendor risk assessment guide](https://superintelligencecompliance.ai/guides/ai-vendor-risk-assessment) covers the checks, and the [AI impact assessment guide](https://superintelligencecompliance.ai/guides/ai-impact-assessment) compares the DPIA with other assessments.

Data sent to a model provider outside the EEA needs a Chapter V mechanism. Implementing Decision (EU) 2023/1795 of July 10, 2023 found an adequate level of protection for personal data transferred to US organisations on the Data Privacy Framework List. The General Court dismissed an action against it on September 3, 2025 (*Latombe v Commission*, T-553/23); an appeal, C-703/25 P, was brought on October 31, 2025, and EUR-Lex listed no Court of Justice judgment on it as of November 12, 2026.

## What have regulators done about AI under the GDPR?

**Selected GDPR enforcement involving AI and automated processing, as of November 12, 2026 (not a complete list)**

| Authority and company | Decision | Amount | Appeal status |
|---|---|---|---|
| Garante (Italy) v. OpenAI | Decision No. 755 of November 2, 2024, announced December 20, 2024: failure to notify a March 2023 data breach; processing users’ data to train ChatGPT without first identifying an adequate legal basis; breach of transparency and information duties; no age verification. Six-month information campaign ordered. | €15 million | Court of Rome judgment No. 4153/2026, published March 18, 2026, upheld the opposition lodged against the decision, and the Garante temporarily removed the decision from its website. The Garante’s pages do not say whether it appealed further. |
| Autoriteit Persoonsgegevens (Netherlands) v. Clearview AI | September 3, 2024: database of scraped facial images and biometric codes that should never have been built; insufficient transparency; no cooperation with access requests. Order to stop the violations. | €30.5 million, plus penalties for non-compliance of up to €5.1 million in total | The AP says Clearview did not object to the decision and is therefore unable to appeal the fine. |
| Autoriteit Persoonsgegevens (Netherlands) v. Uber | August 21, 2026: drivers’ accounts deactivated by fully automated decisions, without human assessment, from 2018 to 2022, in breach of Article 22; drivers not sufficiently informed (Article 13). Case opened on complaints by 171 French drivers, handled under the one-stop-shop procedure. | €824,990,000 | Uber has filed an appeal; the EDPB summary says there is no final judicial decision yet. |

## How does the GDPR fit with the EU AI Act?

Article 2(7) of the AI Act, as amended, says Union law on the protection of personal data, privacy and the confidentiality of communications applies to personal data processed in connection with the rights and obligations the Act lays down and that, without prejudice to Articles 4a and 59, the Act does not affect the GDPR. Deployers of high-risk systems must, where applicable, use the provider’s Article 13 information to carry out their DPIA (AI Act Article 26(9)), and a fundamental rights impact assessment may cross-reference or include parts of a DPIA that already meets an obligation (Article 27(4)).

## Would the Digital Omnibus change the GDPR for AI?

Not unless Parliament and Council adopt it. The Commission’s proposal, COM(2025) 837 of November 19, 2025 (procedure 2025/0360(COD)), is separate from the Digital Omnibus on AI that became Regulation (EU) 2026/1744. Its Article 3 would make the amendments in the table, among other changes.

**Selected GDPR amendments in the Digital Omnibus proposal, COM(2025) 837, and in the Council’s negotiating mandate, none adopted as of November 12, 2026**

| GDPR provision | What the proposal would do if adopted | Council negotiating mandate, October 21, 2026 |
|---|---|---|
| Article 4(1): personal data | Add that information is not personal data for an entity that cannot identify the person, taking into account the means reasonably likely to be used by that entity | Deleted; the current definition is kept |
| Article 9(2)(k) and new Article 9(5): special categories | Add an exception for processing in the context of the development and operation of an AI system or model, with a new Article 9(5) requiring measures to avoid collecting such data and its removal, or protection where removal requires disproportionate effort | Point (k) narrowed to incidental processing of residual data, where not intended by the controller, in the development and technical operation of an AI system or model; such data must be erased without undue delay, or effectively protected where erasure proves technically impossible or requires manifestly disproportionate effort |
| New Article 88c: legitimate interest for AI | Add a provision under which, where processing of personal data is necessary for the interests of the controller in the context of the development and operation of an AI system or model, “such processing may be pursued for legitimate interests within the meaning of Article 6(1)(f) of Regulation (EU) 2016/679, where appropriate, except where other Union or national laws explicitly require consent, and where such interests are overridden by the interests, or fundamental rights and freedoms of the data subject which require protection of personal data, in particular where the data subject is a child” (the proposal’s wording), with safeguards that include an unconditional right to object | Renumbered Article 88 bis: such processing “may be carried out for a legitimate interest of the controller or a third party in accordance with Article 6(1)(f) of this Regulation”; this does not affect the controller’s obligation to choose an appropriate lawful ground, and safeguards, including the Article 21 right to object, apply. The consent carve-out, the overriding-interests clause and the unconditional right to object are deleted |
| Article 22(1) and (2): automated decisions | Rewrite them so a solely automated decision with legal or similarly significant effects may be taken only under one of three conditions, the contract condition applying “regardless of whether the decision could be taken otherwise than by solely automated means.” | Keeps the person’s right not to be subject to such a decision unless one of the three exceptions applies, and deletes the “regardless of whether” clause |

In Joint Opinion 2/2026 (adopted February 10, 2026), the EDPB and EDPS strongly urged the co-legislators not to adopt the change to the definition of personal data, said a specific legitimate-interest provision for AI does not appear necessary given Opinion 28/2024, and said the prohibition in principle on automated decisions should be retained. The Council’s negotiating mandate of October 21, 2026 departs from the proposal on each row of the table. In Parliament, the joint lead committees published a draft report on June 22, 2026; as of November 12, 2026, the Legislative Observatory listed the file as awaiting committee decision.

## How does Switzerland regulate AI and personal data?

Switzerland applies its revised Federal Act on Data Protection (FADP, SR 235.1), in force since September 1, 2023, to circumstances that have an effect in Switzerland, even if initiated abroad (Article 3(1)).

- **FDPIC position.** The Federal Data Protection and Information Commissioner said on November 9, 2023, and repeated on May 8, 2025, that the FADP is directly applicable to AI-supported data processing. In its view, manufacturers, providers and users must make the purpose, functionality and data sources of AI processing transparent; users of language models that communicate directly have a right to know whether they are dealing with a machine; and high-risk cases require a DPIA.
- **Automated decisions (FADP Article 21).** The controller must inform the person of any decision based exclusively on automated processing that has a legal consequence for, or a considerable adverse effect on, them and, on request, let them express their point of view and have the decision reviewed by a natural person. This does not apply where the decision is directly connected with concluding or performing a contract and the person’s request is granted, or where the person explicitly consented to automation. A private person who wilfully fails to inform faces, on complaint, a criminal fine of up to CHF 250,000 (Article 60).
- **AI-specific law.** On February 12, 2025 the Federal Council decided that Switzerland intends to ratify the Council of Europe Convention on Artificial Intelligence and to make the necessary amendments to Swiss law. Switzerland signed the Convention on March 27, 2025. OFCOM’s AI page, published October 28, 2026, says Switzerland does not yet have overarching legislation specifically on AI and that a draft act on transparency, data protection, non-discrimination, supervision and risk assessment is being prepared, with consultation scheduled for spring 2027.
For the UK GDPR, see the [UK AI regulation guide](https://superintelligencecompliance.ai/guides/uk-ai-regulation).

## GDPR AI compliance checklist

- Map each AI system to the personal data it uses, with the purpose of each use, and run a DPIA before any processing likely to result in a high risk
- Document a lawful basis for each purpose; for legitimate interest, keep a three-step assessment covering reasonable expectations and mitigating measures
- Document extraction and query testing before treating a model as anonymous, as EDPB Opinion 28/2024 describes
- Before deploying a third-party model, make and record an appropriate assessment of whether it was developed by unlawfully processing personal data
- Screen training data for special-category data, and rely on AI Act Article 4a only for bias detection and correction, where strictly necessary, and when every condition is met
- Update Article 13 and 14 notices; if you rely on Article 14(5)(b) for scraped data, record why its requirements are fully met and take appropriate measures, including making the information publicly available
- Test whether any output, including a credit score a third party draws strongly on, is an Article 22 decision; if so, fit it within an Article 22(2) exception, provide the Article 22(3) safeguards where the contract or consent exception is used, and be ready to explain the procedure and principles actually applied
- Where an AI vendor processes personal data on your behalf, put the Article 28(3) terms in the contract, and add a Chapter V transfer mechanism where the data leaves the EEA
- Where FADP Article 21 applies, inform the person and, on request, let them be heard and have a natural person review the decision
The [EU AI Act compliance guide](https://superintelligencecompliance.ai/guides/eu-ai-act-compliance) is the hub for EU AI duties.

This guide is general information as of November 12, 2026, not legal advice. Confirm the current text with the competent data protection authority or qualified counsel before acting.

## Sources

- [Regulation (EU) 2016/679 (General Data Protection Regulation)](https://eur-lex.europa.eu/eli/reg/2016/679/oj), Official Journal of the European Union, May 4, 2016.
- [Regulation (EU) 2016/679, document information (amendments and proposed amendments)](https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX:32016R0679), EUR-Lex, accessed November 12, 2026.
- [Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models](https://www.edpb.europa.eu/documents/opinion-of-the-board-art-64/opinion-282024-on-certain-data-protection-aspects-related-to_en), European Data Protection Board, adopted December 17, 2024.
- [Guidelines 02/2026 on Anonymisation, version 1.0 for public consultation](https://www.edpb.europa.eu/documents/guideline/guidelines-022026-on-anonymisation_en), European Data Protection Board, adopted July 7, 2026.
- [Judgment of the Court of Justice, Case C-634/21, *OQ v Land Hessen* (SCHUFA Holding AG intervening)](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62021CJ0634), Court of Justice of the European Union, December 7, 2023.
- [Judgment of the Court of Justice, Case C-203/22, *CK v Magistrat der Stadt Wien* (Dun & Bradstreet Austria GmbH, other party)](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62022CJ0203), Court of Justice of the European Union, February 27, 2025.
- [Regulation (EU) 2024/1689 (Artificial Intelligence Act), consolidated text as of July 27, 2026](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02024R1689-20260727), EUR-Lex, accessed November 12, 2026.
- [Regulation (EU) 2026/1744 (Digital Omnibus on AI)](https://eur-lex.europa.eu/eli/reg/2026/1744/oj?locale=en), Official Journal of the European Union, July 24, 2026.
- [Commission Implementing Decision (EU) 2023/1795 on the adequate level of protection of personal data under the EU-US Data Privacy Framework](https://eur-lex.europa.eu/eli/dec_impl/2023/1795/oj), Official Journal of the European Union, July 10, 2023.
- [Judgment of the General Court, Case T-553/23, *Latombe v Commission*](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62023TJ0553), General Court of the European Union, September 3, 2025.
- [Case C-703/25 P: Appeal brought on 31 October 2025 by Philippe Latombe](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:62025CN0703), Official Journal of the European Union, accessed November 12, 2026.
- [ChatGPT, il Garante privacy chiude l’istruttoria (press release, with a note on Court of Rome judgment No. 4153/2026)](https://www.garanteprivacy.it/home/docweb/-/docweb-display/docweb/10085432), Garante per la protezione dei dati personali, December 20, 2024, accessed November 12, 2026.
- [Dutch DPA imposes a fine on Clearview because of illegal data collection for facial recognition](https://www.autoriteitpersoonsgegevens.nl/en/current/dutch-dpa-imposes-a-fine-on-clearview-because-of-illegal-data-collection-for-facial-recognition), Autoriteit Persoonsgegevens, September 3, 2024.
- [Uber fined nearly 825 million euros for automated driver blocking](https://www.autoriteitpersoonsgegevens.nl/en/current/uber-fined-nearly-825-million-euros-for-automated-driver-blocking), Autoriteit Persoonsgegevens, August 21, 2026.
- [Dutch DPA fines Uber EUR 824 990 000 for unlawful automated decision-making and insufficient information on profiling](https://www.edpb.europa.eu/news/dutch-dpa-fines-uber-eur-824-990-000-for-unlawful-automated-decision-making-and-insufficient_en), European Data Protection Board, October 8, 2026.
- [Proposal for a Regulation as regards the simplification of the digital legislative framework (Digital Omnibus), COM(2025) 837 final](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52025PC0837), European Commission, November 19, 2025.
- [Procedure 2025/0360/COD](https://eur-lex.europa.eu/procedure/EN/2025_360), EUR-Lex, accessed November 12, 2026.
- [Council document 14756/26: Digital Omnibus, mandate for negotiations with the European Parliament](https://data.consilium.europa.eu/doc/document/ST-14756-2026-INIT/en/pdf), Council of the European Union, October 22, 2026.
- [Procedure file 2025/0360(COD)](https://oeil.europarl.europa.eu/oeil/en/procedure-file?reference=2025/0360(COD)), European Parliament Legislative Observatory, accessed November 12, 2026.
- [EDPB-EDPS Joint Opinion 2/2026 on the Proposal for a Regulation as regards the simplification of the digital legislative framework (Digital Omnibus)](https://www.edpb.europa.eu/documents/legislative-opinion/edpb-edps-joint-opinion-22026-on-the-proposal-for-a-regulation-as_en), European Data Protection Board and European Data Protection Supervisor, February 10, 2026.
- [Federal Act on Data Protection (FADP), SR 235.1 (English translation, no legal force)](https://www.fedlex.admin.ch/eli/cc/2022/491/en), Fedlex, accessed November 12, 2026.
- [Update: Current data protection legislation is directly applicable to AI](https://www.edoeb.admin.ch/en/update-current-legislation-directly-applicable-ai), Federal Data Protection and Information Commissioner, May 8, 2025.
- [AI regulation: Federal Council to ratify Council of Europe Convention](https://www.bakom.admin.ch/en/nsb?id=104110), Swiss Federal Council, February 12, 2025.
- [Switzerland Signs the Council of Europe’s Global Treaty on AI](https://www.coe.int/en/web/artificial-intelligence/-/switzerland-signs-the-council-of-europe-s-global-treaty-on-ai), Council of Europe, March 27, 2025.
- [Artificial Intelligence](https://www.bakom.admin.ch/en/artificial-intelligence), Federal Office of Communications (OFCOM), October 28, 2026.

## Common questions

**Does the GDPR apply to AI models?**

Within its scope, it applies whenever personal data is processed to build, train, test or run an AI system, and, in the EDPB's view, to a model trained on personal data unless the model is anonymous. EDPB Opinion 28/2024 says models trained with personal data cannot, in all cases, be considered anonymous, and that anonymity claims should be assessed by supervisory authorities case by case.

**Can a company use legitimate interest to train AI under the GDPR?**

It can rely on Article 6(1)(f) only if the processing passes the three-step test the EDPB recalls in Opinion 28/2024: an interest that is lawful, clearly and precisely articulated, and real and present; necessity, meaning the processing allows the interest to be pursued with no less intrusive way to pursue it; and a balancing test that the interest is not overridden by data subjects' interests or fundamental rights and freedoms, taking account of their reasonable expectations. Public authorities cannot use point (f) in the performance of their tasks. Special-category data also needs an Article 9 exception.

**Has the Digital Omnibus changed the GDPR for AI?**

No. The Commission's November 19, 2025 proposal, COM(2025) 837, would add AI-specific rules such as an Article 88c on legitimate interest for AI. The Council agreed its negotiating mandate on October 21, 2026, keeping the current definition of personal data, but as of November 12, 2026 Parliament had not adopted its position and no act had been adopted. It is a proposal, not law. The separate Digital Omnibus on AI, Regulation (EU) 2026/1744, amended the AI Act, and as of November 12, 2026 the GDPR's text had not been amended.

