Classify your system under the EU AI Act in 3 minutes.

The EU AI Act, Regulation (EU) 2024/1689, is the first European law to regulate artificial intelligence systems. It applies to anyone placing an AI system on the Union market, or whose system outputs are used in the Union, regardless of where the company is established: providers that develop and market the system, importers that bring it into the EU from a third country, distributors that resell it, and "deployers" that use it in the course of a professional activity. If your company falls into any of those four roles, it applies to you.

The Regulation entered into force on 1 August 2024 with a phased timeline (Art. 113). The prohibited practices in Art. 5 are enforceable from 2 February 2025 and general-purpose AI model obligations (GPAI, Chapter V) from 2 August 2025. From 2 August 2026 the Art. 50 transparency obligations, the Chapter IX market-surveillance regime and the Art. 101 fines apply. Regulation (EU) 2026/1744 (the Digital Omnibus on AI, in force since 27 July 2026) postponed the high-risk regime: Chapter III, Sections 1, 2 and 3 apply from 2 December 2027 for systems under Art. 6(2) and Annex III, and from 2 August 2028 for those under Art. 6(1) and Annex I. The postponement does not change the substance of the obligations.

Most likely yes, in the role of "deployer" (Art. 26). If you use a third-party AI system in the course of a professional activity, the Art. 26 obligations apply depending on the specific use: human oversight, training of staff who operate it, logging of serious incidents and, if the system is used for one of the Annex III high-risk cases (HR, credit scoring, education, etc.), the rest of the high-risk deployer duties. The fundamental rights impact assessment (Art. 27) is stricter and only reaches public bodies, providers of a public service and those using the system for credit or for life and health insurance: a private CV screen does not need one. The model obligations (Art. 53/55) belong to the LLM provider, not to you as integrator.

Three minimum points. One, transparency (Art. 50): you have to inform the user that they are interacting with AI, unless it is obvious. Two, if the chatbot makes decisions that affect the user's rights (access to a service, pricing, hiring), it can fall under Annex III high risk and trigger the entire Chapter III regime (risk management, data governance, technical documentation, human oversight). Three, if you integrate a third-party GPAI (OpenAI, Anthropic, Google), you have to receive and retain the provider's technical documentation (Art. 53(1)(b)) and apply the deployer obligations (Art. 26).

The provider (Art. 3(3)) is the one who develops an AI system or has it developed and markets it under their name or trademark: they carry the bulk of the obligations on design, documentation, conformity assessment and EU registration. The deployer (Art. 3(4), the former "user") is the one who uses the system under their authority in a professional, non-domestic activity: they have lighter obligations on oversight, training and incident logging; and, only if they are a public body, provide a public service or use the system for credit or life and health insurance, the Annex III fundamental rights impact assessment (Art. 27). The same company can be a provider for one system and a deployer for another.

It is high risk if it falls into one of two cases. One, Annex I: a safety component of a product regulated by EU harmonisation legislation (machinery, toys, lifts, medical devices, etc.). Two, Annex III: any of the eight Annex III domains (biometrics, critical infrastructure, education, employment, access to essential services, law enforcement, migration, justice). If it does, you must comply with Chapter III: risk management system, data governance, technical documentation, human oversight, conformity assessment, registration in the EU database, CE marking and post-market monitoring.

Art. 99 sets three tiers, and each takes whichever is higher of the fixed amount and the percentage of total worldwide annual turnover for the preceding financial year. Up to €35M or 7% for using the Art. 5 prohibited practices. Up to €15M or 3% for breaching the rest of the Regulation's obligations (provider, importer, distributor, deployer, transparency and high-risk systems). Up to €7.5M or 1% for giving authorities incorrect information. If you train or develop a general-purpose AI (GPAI) model, that sits under a separate regime: the fine comes from the European Commission, not your national authority, with its own cap of up to €15M or 3% of worldwide turnover (whichever is higher), applying from 2 August 2026 (Art. 101). For SMEs and start-ups the cap is the lower of the two figures, not the higher one (e.g.: if 3% of your turnover exceeds €15M, the cap is €15M); Regulation (EU) 2026/1744 extended that treatment to small mid-caps. And you can only be fined for an obligation that is already applicable: the high-risk regime does not apply until December 2027 (Annex III) or August 2028 (Annex I).

A little, but it does not exempt you. Several of its support measures are aimed squarely at this calculator's audience. Every country must run at least one regulatory sandbox: a supervised environment where you can validate an innovative system before launch, with free, priority access for SMEs and start-ups (Art. 57 and 58). Conformity-assessment fees are reduced in proportion to your size (Art. 62). You can file the Annex IV technical documentation using the simplified form the Commission publishes, and assessment bodies must accept it (Art. 11). The quality-management system can scale to your organisation's size (Art. 17), and a microenterprise with no associated or linked companies can keep that paperwork lighter still (Art. 63). E.g.: a start-up shipping a high-risk system requests a priority sandbox slot and documents with reduced templates. What changes is the support and the cost of the process, not the substantive obligations.

No. Every year the Commission reviews whether to add or remove high-risk Annex III uses and prohibited practices, and it can do so by delegated act (Art. 7 and 112). A "you're not in Annex III" today can be out of date tomorrow. E.g.: your system is outside the high-risk regime now, but a new delegated act could bring its domain in. It is worth re-classifying whenever your use case changes or the Regulation is updated.

Because the technical standards you would demonstrate conformity against are not published yet. Regulation (EU) 2026/1744 pushed the timeline back until those harmonised standards and common specifications appear (Art. 40 and 41): December 2027 for Annex III and August 2028 for Annex I. E.g.: you can't "CE-mark" your HR system today because the harmonised standard to audit it against doesn't exist yet. The obligation still stands; only the date moved, not the substance.

Two things worth knowing up front. One, the database is public: anyone can look up that your system exists and who offers it (Art. 49 and 71). E.g.: a CV-screening system will appear in the register under its provider's name. Two, registering is more than signing up: it will ask you to identify the system, describe its purpose and the data it uses, and attach the EU declaration of conformity and the certificate if a notified body was involved (Annex VIII). E.g.: have the purpose summary ready before you open the file.

It doesn't take an inspection for this to matter. From 2 August 2026, anyone who believes an AI system breaches the Regulation can lodge a complaint with the market surveillance authority (Art. 85): a customer, a competitor or a former employee. And if you make decisions about people with a high-risk system, those people can ask you for a clear explanation of the AI's role (Art. 86). Keeping your documentation in order is your best defence.

It is a screening tool. You answer a few questions about your system and it returns the risk level under Regulation (EU) 2024/1689, the articles that apply to you and the date they become enforceable, already on the Digital Omnibus timeline. That covers the first step, classification, which is where most teams get stuck. It is not compliance software: it does not build the Art. 9 risk management system, the Annex IV technical documentation or the EU database registration. That takes a project, not a form. If the result comes back high risk, the tool tells you which obligations open up and from when.

No. The calculator applies the decision tree of Regulation (EU) 2024/1689 to your answers and returns the corresponding risk level and planes. A final classification requires a specific analysis of the system and a legal reviewer. Treat it as a starting point, not a substitute for advice.

The decision tree is built on top of the consolidated text of Regulation (EU) 2024/1689 published on EUR-Lex. Each article cited in the result links directly to EUR-Lex in the page's language. For practical interpretation, the official reference is the guidance the European Commission publishes on how to apply the Regulation (Art. 96).

The Regulation sorts AI systems into four risk levels: prohibited (Art. 5), high risk (Art. 6 with Annexes I and III), transparency (Art. 50) and minimal. On top of that sit two planes that are not risk levels but change your obligations: whether you build a general-purpose model (GPAI, Chapter V), which can also carry systemic risk (Art. 51 and 55); and your role (provider, deployer, distributor or importer). That is why the result gives you all three at once: risk level, whether a GPAI model is involved and the role you fall into.

Yes, one light and cross-cutting: AI literacy (Art. 4). If you are a provider or you use AI in your professional activity (deployer), you have to take measures so your team and anyone operating the system on your behalf understand the AI they handle, adapted to their training and the context of use. It does not require a certificate or a specific level: it is an obligation of means, applicable from 2 February 2025. E.g. a short internal briefing on what the tool does, its limits and when to distrust its output is usually enough for low-risk use.

Yes. After classifying there is a Download as PDF button that generates an A4 document with the classification, the cited articles, the applicable deadline, the obligations to map and the next actions.