An AI company entering the EU must first decide which role it holds under the AI Act, because that role sets its obligations. A non-EU provider generally needs an authorized representative in the Union, and GDPR can require a separate EU representative under Article 27. As of November 2026, most AI Act obligations apply, while high-risk rules start on 2 December 2027 and 2 August 2028.
Start with scope, not with a checklist
Regulation (EU) 2024/1689, the AI Act, reaches well beyond companies headquartered in Europe. Article 2(1) on EUR-Lex applies it to providers that place AI systems on the market or put them into service in the Union, whether or not the provider is established there. It also applies to deployers located in the Union, to providers and deployers in third countries where the output of the AI system is used in the Union, and to importers, distributors and authorized representatives.
The output clause is the one that surprises people. A company with no EU office, no EU sales team and no EU customer contract can still fall within scope if the output of its system is used in the Union. Before you plan a launch, map where your outputs are used, not only where your customers are incorporated.
The Act also contains exclusions. Article 2 does not apply to systems used exclusively for military, defense or national security purposes, to systems developed and put into service for the sole purpose of scientific research, or to research and testing before a system is placed on the market, with testing in real-world conditions not covered by that last exclusion. Check each one against your facts rather than assuming it applies.
The five roles that matter
Article 3 of the AI Act defines the roles, and the Commission's AI Act Service Desk publishes the text. Read the definitions closely, because the labels are narrower than they sound.
- Provider: a person or body that develops an AI system or general-purpose AI model, or has one developed, and places it on the market or puts it into service under its own name or trademark, whether for payment or free of charge.
- Deployer: a person or body using an AI system under its authority, except for personal non-professional use.
- Importer: a person or body located or established in the Union that places on the market an AI system bearing the name or trademark of a person established in a third country.
- Distributor: a person in the supply chain, other than the provider or importer, that makes an AI system available on the Union market.
- Authorized representative: a person or body located or established in the Union that has received and accepted a written mandate from a provider to carry out the provider's obligations and procedures under the Regulation.
One company can hold several roles at once. A U.S. software company that builds a model and sells it under its own brand is a provider. If it also runs the system for its own EU employees, it is a deployer of that system too.
Roles can also shift. Article 25 treats a distributor, importer, deployer or other third party as the provider of a high-risk system if they put their name or trademark on a system already on the market, make a substantial modification that leaves it high-risk, or change the intended purpose of a system so that it becomes high-risk. A reseller that rebrands your product therefore inherits your obligations, which is a reason to control the reseller contract.
What each role has to do
The heaviest duties fall on providers of high-risk systems, and the obligations for other roles are narrower but real.
Article 22 requires a provider established in a third country to appoint, by written mandate, an authorized representative established in the Union before making a high-risk system available on the Union market. The mandate must give the representative authority to verify that the EU declaration of conformity and technical documentation exist and that a conformity assessment was done, to keep that material and the provider's contact details available to authorities for 10 years, to provide documentation and log access on a reasoned request, to cooperate with authorities, and to handle registration where applicable. The representative must end the mandate and inform the authority if it believes the provider is acting contrary to its obligations.
Article 54 sets a similar duty for providers of general-purpose AI models established in third countries, before they place a model on the Union market. It exempts providers of models released under a free and open-source license with publicly available weights, unless the model presents systemic risk.
Importers must verify conformity before placing a high-risk system on the market and put their contact details on it. Distributors must check for the CE marking, the EU declaration of conformity and the instructions for use. Deployers of high-risk systems must use them according to the instructions, assign human oversight to people with the necessary competence, training and authority, monitor operation, and keep automatically generated logs for at least six months, unless other law says otherwise.
The penalties are set in Article 99. Breaching the prohibited-practices rules can reach EUR 35 million or 7% of worldwide annual turnover, whichever is higher, and breaching the listed operator obligations can reach EUR 15 million or 3%. Supplying incorrect or misleading information to authorities can reach EUR 7.5 million or 1%.
GDPR runs alongside the AI Act
The AI Act does not replace GDPR. Article 2(7) of the AI Act says the Act does not affect Regulation (EU) 2016/679, and AI products that process personal data face both regimes.
GDPR Article 3 on EUR-Lex has two triggers for companies outside the EU. Article 3(1) covers processing in the context of an establishment in the Union. Article 3(2) covers processing of personal data of people in the Union by a controller or processor not established there, where the processing relates to offering goods or services to those people, or to monitoring their behavior within the Union.
Where Article 3(2) applies, Article 27 requires the controller or processor to designate a representative in the Union in writing. The representative must be established in a Member State where the affected people are, and must be mandated to be addressed, in addition to or instead of the controller or processor, by supervisory authorities and data subjects. The duty does not apply to occasional processing that avoids large-scale special-category data and is unlikely to risk individuals' rights, or to public authorities.
These are two separate appointments with two separate mandates. The AI Act representative answers to market surveillance authorities about conformity. The GDPR representative answers to data protection authorities and data subjects. The same firm can hold both roles, but each mandate has to be drafted for its own purpose.
The European Data Protection Board's Opinion 28/2024, adopted on 18 December 2024, addresses AI models directly. It says an AI model trained on personal data cannot be considered anonymous in every case, and that anonymity claims should be assessed case by case. For a model to count as anonymous, the likelihood of extracting personal data from it, and of obtaining such data from queries, must be insignificant. The Opinion also recalls the three-step test for legitimate interest: identify the interest, test necessity, and balance it against data subjects' rights.
For anyone deploying a third-party model, the Opinion adds a duty. Where personal data is retained in a model and processed by another controller, supervisory authorities should consider whether that deploying controller assessed, as part of its accountability obligations, that the model was not developed through unlawful processing.
The timeline as of November 2026
The European Commission's AI Act page gives the current dates. The Act entered into force on 1 August 2024 and became applicable on 2 August 2026, with exceptions.
- 2 February 2025: prohibited practices and AI literacy obligations applied.
- 2 August 2025: governance rules and obligations for general-purpose AI models applied.
- 2 August 2026: general application, including the transparency rules, and the start of enforcement by the AI Office and Member State authorities.
- 2 December 2026: a prohibition on AI systems that generate non-consensual sexually explicit content or child sexual abuse material takes effect, added by the AI Omnibus.
- 2 December 2027: obligations for high-risk systems in the Annex III areas, such as biometrics, critical infrastructure, education, employment, migration and border control.
- 2 August 2028: obligations for high-risk systems embedded in regulated products under Annex I.
The consolidated text adds two transition rules in Article 111. Providers of general-purpose AI models placed on the market before 2 August 2025 must comply by 2 August 2027. Providers of generative systems placed on the market before 2 August 2026 must comply with the Article 50(2) marking duty by 2 December 2026.
What the Digital Omnibus changed
In late 2025 the Commission proposed changes to AI Act timing under its Digital Package on Simplification. The AI part, called the AI Omnibus, was proposed on 19 November 2025. The Commission reports a political agreement on 7 May 2026 and entry into force on 27 July 2026. It was published as Regulation (EU) 2026/1744.
Its main effect is on dates. High-risk obligations that originally applied from 2 August 2026 for Annex III systems and 2 August 2027 for Annex I systems now apply from fixed later dates, 2 December 2027 and 2 August 2028. The Commission says it also extends certain simplified requirements for small and medium-sized enterprises to small mid-cap companies, expands regulatory sandbox access, reinforces the AI Office's powers over systems built on general-purpose models, and adds the new prohibition.
The delay is not a repeal. The high-risk requirements remain, including risk management, data quality, logging, documentation, human oversight, and accuracy and cybersecurity. Do not read the later dates as permission to defer design work, since the controls named in the Act take engineering time to build.
Where to start
Begin with a role and scope map, then build outward from it. These steps fit most AI companies preparing an EU launch.
- List every AI system and model you will offer, and record for each one whether you are provider, deployer, importer or distributor in the EU supply chain.
- Classify each system against Article 5 prohibitions, Annex III and Annex I, and the Article 50 transparency rules that apply from August 2026.
- Decide whether you need an AI Act authorized representative, a GDPR Article 27 representative, or both, and draft each written mandate separately.
- Review reseller, integrator and cloud contracts for the Article 25 triggers, so no partner changes your role by rebranding or modifying a system.
- Prepare a data protection position for each model: the legal basis, the lawfulness of training data, and the evidence for any anonymity claim, in line with Opinion 28/2024.
- Set up the records that representatives and authorities will ask for, including technical documentation, logs and a 10-year retention schedule.
Legal counsel should confirm the role analysis and the mandates for your specific facts. The aim of this work is to arrive in the market with the roles, representatives and records already in place.
References
- EUR-Lex: Regulation (EU) 2024/1689 (Artificial Intelligence Act)
- EUR-Lex: Regulation (EU) 2026/1744 (Digital Omnibus on AI)
- European Commission: AI Act, regulatory framework for AI
- European Commission AI Act Service Desk: Article 22, Authorised representatives of providers of high-risk AI systems
- European Commission AI Act Service Desk: Article 54, Authorised representatives of providers of general-purpose AI models
- EUR-Lex: Regulation (EU) 2016/679 (General Data Protection Regulation)
- European Data Protection Board: Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models





