EU AI Act checker

Find your AI system's likely risk class under the EU AI Act, the obligations that come with your role, and the dates they apply. Free, no login, and your answers stay in your browser.

General information, not legal advice. Based on Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744, the Digital Omnibus on AI. Last verified 19 October 2026.

Does the system reach the EU?

The Act covers providers who supply AI systems in the EU from anywhere, deployers in the EU, and providers or deployers elsewhere whose system's output is used in the EU (Art. 2(1)).

YesYou are in the EU, you supply or use the system there, or its output is used there NoNone of these applies Not sureThe checker treats this as yes

Does any of these describe the system? Used only for military, defence or national securityArt. 2(3) Built and used only for scientific research and developmentArt. 2(6) Still in research, testing or development, not yet released or in serviceTesting in real-world conditions is not excluded, Art. 2(8) Used by individuals only for personal, non-professional purposesArt. 2(10) None of these

What is your role? Select all that apply ProviderYou develop the system, or have it developed, and place it on the market or put it into service under your own name or trademark, paid or free, Art. 3(3) DeployerYou use the system under your authority in a professional activity, Art. 3(4) ImporterYou are in the EU and place on the market a system carrying the name or trademark of a company outside the EU, Art. 3(6) DistributorYou make the system available on the EU market and are neither its provider nor its importer, Art. 3(7)

What does the system do? Select all that apply Talks or chats with peopleA chatbot, voice assistant or support agent, Art. 50(1) Generates synthetic audio, images, video or textArt. 50(2) Creates images, audio or video that could pass as real people, places or eventsDeep fakes, Art. 50(4) Writes text published to inform the public on matters of public interestArt. 50(4) Recognises emotions or intentions from biometric dataFor example from a face or voice, Art. 50(3) and Annex III, point 1(c) Sorts people into categories from biometric dataArt. 50(3) Takes actions through tools or APIs, as an AI agentThe Act has no separate class for agents: an agent is classified by its use None of these

Is it a safety component of a product, or itself a product, under the EU product laws in Annex I?

A safety component protects health and safety, or endangers them if it fails. Features used only for user assistance, performance, efficiency, automation, convenience or quality control do not count (Art. 6(1a) to 6(1c)).

Yes, and the product needs a third-party conformity assessmentToys, lifts, radio equipment, pressure equipment, medical and in vitro diagnostic devices, personal protective equipment, gas appliances, cableways, recreational craft or equipment for explosive atmospheres (Annex I, Section A) Yes, in machinery, vehicles, aviation, marine equipment or rail, with a third-party assessmentAnnex I, Section B Yes, but no third-party conformity assessment is needed No

Do you provide or build on a general-purpose AI model? No We build on a model from another companyFor example through an API We provide a general-purpose AI modelA model able to perform a wide range of tasks, placed on the EU market through an API, a download or open weights, Art. 3(63) We provide a general-purpose AI model with systemic riskTrained with more than 10²⁵ floating-point operations, or designated by the Commission, Art. 51

How the checker works

The questions follow the order the AI Act itself uses. First scope: whether the system reaches the EU and whether an exclusion in Article 2 applies. Then your role, because providers, deployers, importers and distributors carry different duties. Then the prohibited practices in Article 5, the product legislation route to high-risk in Article 6(1) and Annex I, the high-risk uses in Annex III, the transparency duties in Article 50, and the separate rules for general-purpose AI models.

The result names one risk class, lists the obligations for each role you selected with links to the articles on EUR-Lex, and shows the dates that apply. Everything runs in this page: your answers are never sent, and nothing you enter is stored. If you allowed analytics cookies, the site counts that a check finished and which risk class it showed. You can print the result or save it as a PDF from your browser.

Key dates

2 February 2025: prohibited practices and AI literacy (Art. 113, point (a)).
2 August 2025: obligations for providers of general-purpose AI models (Art. 113, point (b)).
2 August 2026: Article 50 transparency obligations (Art. 113, second paragraph).
2 December 2026: the two prohibitions Regulation (EU) 2026/1744 added, on non-consensual intimate content and child sexual abuse material (Art. 113, point (a)).
2 December 2027: high-risk rules for Annex III systems (Art. 113, point (c)(i), as amended by Regulation (EU) 2026/1744).
2 August 2028: high-risk rules for Annex I systems (Art. 113, point (c)(ii), as amended by Regulation (EU) 2026/1744).
2 August 2030: high-risk systems for public authorities that were on the market or in service before the high-risk rules applied must comply (Art. 111(2)).

Sources: Regulation (EU) 2024/1689, consolidated text of 27 July 2026 and Regulation (EU) 2026/1744 of 8 July 2026, which entered into force on 27 July 2026. Both on EUR-Lex, last verified 19 October 2026.

What the result is not

The checker is a starting point, not legal advice. It simplifies some tests, such as what counts as a safety component or a substantial modification, and it does not cover national law, sector rules, data protection or the guidance the Commission publishes. Read the cited articles and take legal advice before you rely on a result.

Asqav signs the agent actions your integration sends, and anyone holding a record can verify it independently. Those records support duties such as logging, human oversight and log retention in Articles 12, 14, 19 and 26, and the EU AI Act audit pack collects them for review. A record supports your compliance work; it does not establish compliance by itself.