The classifier takes answers to eight questions on role, establishment, Annex III match, Annex I product, the Art. 6(3) filter, Art. 50 functions, the Art. 27 deployer condition and general-purpose AI status. The result lists a risk tier, the obligation set for that role and tier with each article's application date from the regulation dataIn force, the artifacts an assessor asks for, and the matched rules. The tool runs in the browser and sends nothing. Every result is a reasoned technical assessment for discussion with counsel.
How the result is built
The form shows only the questions that apply to the role answered. Every organisation answers the role and establishment questions, with the definitions of Art. 3(3) to 3(8) and the Art. 25(3) rule for a product manufacturer in the option labels. A provider, product manufacturer, deployer, importer or distributor answers the Annex III, Annex I and Art. 50 questions; the Art. 6(3) filter appears only after an Annex III match, and the Art. 27(1) deployer question only for a deployer with an Annex III match. A general-purpose AI model provider answers the Art. 51(2) compute question instead, where the presumption of high-impact capabilities turns on cumulative training compute above 10^25 floating point operations or on a Commission designation under Art. 51(1)(b).
Every rule in the classifier logic is evaluated against the answers. A rule that matches adds a tier candidate, one or more obligation profiles from the profiles logic, notes, and the id of any unsettled point it names. The result tier is the first candidate in the profiles' tier order: Annex III high-risk under Art. 6(2); Annex I Section A high-risk under Art. 6(1); Annex I Section B product under Art. 2(2); Annex III area not high-risk under Art. 6(3); general-purpose AI model with systemic risk under Art. 51 and Art. 55; general-purpose AI model under Art. 53 and Art. 54; Art. 50 transparency duty only; and no obligation identified within the tool's scope, where Art. 4 on AI literacy still applies. The Art. 4 profile matches every role. The matched rule ids are listed under the result and link to the rule on the logic page.
Each profile row names a regulation entry, and the "Applies from" column renders that entry's date and status from the same data the regulation page renders. An article id with an obligation page (Art. 9 to Art. 17, Art. 26, Art. 27, Art. 50, Art. 53 to Art. 55) resolves to its record in the obligation data: the row shows the obligation title and the services that test it, and the artifact list under "Artifacts an assessor asks for" is the record's artifact list. An article without an obligation page, such as Art. 16, Art. 22, Art. 23, Art. 24, Art. 43, Art. 49 or Art. 72, carries a label and a date only. The question set follows the determinations of PR-CLS v1.0, which runs the same checks against the intended-purpose materials and the observed behaviour of the system rather than against self-reported answers; the engine and the version of every logic file are described on the methodology page.
What the result is not
The result is not a classification decision under Art. 6. Where a provider concludes that an Annex III system is not high-risk, the assessment under Art. 6(4), its documentation and the registration under Art. 49(2) stay with the provider; the tool records none of them. The result is not a conformity assessment under Art. 43, and it is not legal advice; the disclaimer applies to every result. The tool does not assess the Art. 5 prohibited practices, the Art. 25(1) triggers that turn a deployer or distributor into a provider, or sector rules outside the Act. A market surveillance authority, the AI Office or a court can reach a different result. Three rules name an unsettled point; where such a rule matches, the result shows both readings, and the same three points follow.
Where the output goes next
The result panel carries a share link that encodes the answers in the page address, and three "Open in" links that pass the role, tier, establishment and Art. 27 answer forward. The obligation tracker lists each article of the profile with its regulation entry; the test checklist generator lists the testable, documentation and process elements of each obligation record for the role and tier; the traceability matrix template returns a requirement-to-evidence matrix for download. The outlines on the templates page give the Annex IV, FRIA, protocol and evidence pack structures those documents refer to. Where the answers are to be verified against the intended-purpose materials and the observed behaviour of the system rather than self-reported, the classification and readiness assessment service runs PR-CLS v1.0 and produces the classification memo and the obligation register; the limits of every determination are stated on what we cannot do.
Questions
+−How does the EU AI Act risk classification work?
Art. 6(1) of Regulation (EU) 2024/1689 classifies as high-risk an AI system that is a safety component of, or is itself, an Annex I Section A product that requires third-party conformity assessment. Art. 6(2) classifies as high-risk a system referred to in Annex III. Art. 6(3) removes an Annex III system that performs a narrow procedural task, improves a completed human activity, detects decision patterns without replacing human assessment, or performs a preparatory task. Profiling of natural persons keeps a system high-risk under Art. 6(3).
+−EU AI Act risk classification: is your AI high-risk?
An AI system is high-risk under Art. 6(2) where its intended purpose falls within an Annex III area, such as employment, education, essential services or biometrics, and no Art. 6(3) condition applies. An AI system is high-risk under Art. 6(1) where it is a safety component of an Annex I Section A product that needs third-party conformity assessment. The classifier on this page returns the tier for a set of answers as a technical assessment for discussion with counsel; the assessment under Art. 6(4) remains the provider's own.
+−What is an EU AI Act risk classification assessment?
A risk classification assessment determines the role of the organisation under Art. 3, whether the system is high-risk under Art. 6 with Annex I or Annex III, whether an Art. 6(3) condition applies, which Art. 50 paragraphs apply to each function, and whether a model is a general-purpose AI model under Art. 3(63). The classification and readiness assessment service runs the assessment under PR-CLS against the intended-purpose materials and the observed behaviour of the system, and issues a classification memo and an obligation register.
+−How to classify AI systems under the EU AI Act?
Classification runs in a fixed order. Determine the role under Art. 3(3) to 3(8), with Art. 25(3) for a product manufacturer. Check whether the system is, or is a safety component of, an Annex I product under Art. 6(1). Check whether the intended purpose matches an Annex III area under Art. 6(2). Apply the Art. 6(3) conditions and the profiling rule. Check which Art. 50 paragraphs apply to each function. For a general-purpose AI model, compare cumulative training compute with the Art. 51(2) presumption of 10^25 floating point operations.
+−Is there an AI Act self-assessment tool?
The classifier on this page is a self-assessment tool: the organisation answers the questions itself, and the rules run in the browser from data bundled at build. Every result is labelled a reasoned technical assessment for discussion with counsel. Under Art. 6(4), a provider that concludes an Annex III system is not high-risk documents that assessment before placing the system on the market; the classifier records neither the documentation nor the Art. 49(2) registration. The obligation tracker, the checklist generator and the traceability matrix expand the result.
Every inquiry receives a reply within 24 hours. Scope and availability are discussed per request.