Which EU AI Act risk category is your AI system in?
Answer about 15 questions about your AI system. See its likely risk category under the EU AI Act, whether you are its provider or deployer, and a checklist of what applies, with articles and dates. Free, no account, about 5 minutes. Not legal advice: a first screening to discuss with a lawyer.
How it works
- You answer questions that follow the AI Act's own tests: where the system is used, your role, the banned practices, the two high-risk routes and the transparency duties.
- Our rules apply the Act in its own order: scope, prohibited practices, Annex I, Annex III and the Article 6(3) exception, then transparency. No AI is involved, and the same answers always give the same result.
- You get a checklist: the likely category, your likely role, why we think so (with links to change any answer) and each obligation with its article and the date it applies from.
- Take it further if you want: get it by email, discuss it with a lawyer, or carry the answers into an AI-BOM that documents the system.
The four risk categories, in plain words
| Category | What it means | Examples |
|---|---|---|
| Prohibited | Banned in the EU since 2 February 2025 (Article 5). | Social scoring, emotion recognition at work or school, manipulative AI causing significant harm. |
| High risk | Allowed, with the most obligations: risk management, documentation, logs, human oversight, conformity assessment (Articles 6 to 49). | CV screening, credit scoring, exam grading, AI safety components in medical devices or machinery. |
| Transparency | Allowed; people must know they deal with AI or see AI-made content (Article 50). | Chatbots, image and text generators, deepfakes. |
| Minimal | No specific obligations beyond AI literacy (Article 4); voluntary codes of conduct. | Spam filters, search, recommendations, code assistants, most internal tools. |
Provider or deployer?
The provider develops an AI system (or has it developed) and offers it under its own name. The deployer uses an AI system in its work. If your product calls an AI provider's API, you are usually the provider of your product, and the model company is the provider of the general-purpose model. A deployer who puts its name on a high-risk system, changes it substantially or changes what it is used for becomes its provider (Article 25). More on roles.
What counts as high risk
Two routes. An AI that is a safety component of a product under EU product safety law that needs a third-party conformity assessment (Annex I: machinery, toys, medical devices, vehicles…). Or an AI used for one of the purposes listed in Annex III:
- Biometrics: remote biometric identification (not just confirming that someone is who they claim to be), biometric categorisation by sensitive traits, or emotion recognition.
- Critical infrastructure: a safety component in managing or operating critical digital infrastructure, road traffic, or the supply of water, gas, heating or electricity.
- Education and vocational training: deciding access or admission, evaluating learning outcomes, assessing the level of education someone will get, or monitoring students for cheating during tests.
- Employment and workers management: recruiting or selecting people (targeted job ads, filtering applications, evaluating candidates), or making decisions on promotion, termination, task allocation based on behaviour or traits, or monitoring and evaluating workers' performance and behaviour.
- Essential services, credit and insurance: deciding eligibility for public benefits and services, evaluating creditworthiness or credit scores (except to detect fraud), risk assessment and pricing for life or health insurance, or evaluating and prioritising emergency calls and first response.
- Law enforcement: use by or for law enforcement: assessing the risk of someone becoming a victim, polygraphs, evaluating the reliability of evidence, assessing the risk of offending (not only by profiling), or profiling in investigations.
- Migration, asylum and border control: use by or for public authorities: polygraphs, assessing security, health or irregular migration risks, examining applications for asylum, visas or residence permits, or detecting and identifying people at borders.
- Justice and democratic processes: helping a judicial authority research and interpret facts and the law and apply it, or influencing the outcome of an election or referendum, or how people vote.
An Annex III system escapes the high-risk rules only when it does not pose a significant risk, for example when it only does a narrow procedural task, and never when it profiles people (Article 6(3)).
If you only call an AI provider's API
Using a model through an API does not make you a general-purpose AI model provider. Your obligations come from what your product does: a writing assistant is likely minimal risk with transparency duties for generated content; a tool that ranks job candidates is high risk, whoever trained the model. The model company's own duties (Articles 53 to 55) do not move to you, but you should keep track of which model and version you use.
Key dates
| 1 August 2024 | The AI Act enters into force. |
| 2 February 2025 | Prohibited practices are banned; AI literacy duties start (Articles 4 and 5). |
| 2 August 2025 | Obligations for general-purpose AI models apply to new models (Articles 53 to 55); penalties apply. |
| 2 August 2026 | Most of the Act applies: high-risk systems in Annex III areas, transparency duties (Article 50), deployer obligations. |
| 2 August 2027 | High-risk rules apply to AI in products covered by Annex I law (Article 6(1)); general-purpose models placed on the market before 2 August 2025 must comply. |
Dates checked against Regulation (EU) 2024/1689 (the AI Act) on 17 November 2026. In November 2025 the European Commission proposed a "Digital Omnibus" that would postpone the high-risk obligations, linking them to the availability of standards (at the latest December 2027 for Annex III areas and August 2028 for Annex I products). Check whether it has been adopted, and its final dates, before relying on the high-risk dates shown here.
What to document
Whatever the category, write down what the system is for, who it affects, which models, data and AI services it uses, and how people oversee it. That is what an AI-BOM holds: AIssue finds the AI components in your code and asks you the rest, and your answers here carry over.
Guides
Questions
Is this legal advice?
No. It is a first screening, based only on your answers, to help you prepare a conversation with a lawyer. Every result cites the articles of the AI Act it relies on, and says when an answer needs a closer look.
Do you keep my answers?
No, unless you ask for the checklist by email. Your answers are kept in the result link itself (signed, so it cannot be changed), which works for 30 days; anyone with the link can see them. The description of your system is never put in the link. If you ask for the email, the answers are kept for 90 days, then deleted.
My product just calls an AI provider's API. Does the AI Act apply to me?
Usually yes, as the provider of your own AI system (your product), while the model company is the provider of the general-purpose model. What you must do depends on what your product is used for: most uses are minimal risk or only have transparency duties.
What makes an AI system high risk?
Two routes: it is a safety component of a product covered by EU product safety law (Annex I), or it is used for one of the purposes listed in Annex III, such as hiring, credit scoring, education, essential services, law enforcement, migration or justice.
When do the obligations apply?
In stages: prohibitions since 2 February 2025, general-purpose AI model rules since 2 August 2025, most of the rest from 2 August 2026, and high-risk AI in regulated products from 2 August 2027. Proposals to postpone the high-risk dates exist: the result shows when we last checked the dates.
What happens with my answers if I create an account?
The result link carries them to your first AI-BOM, where AIssue offers to fill in the matching answers (who is affected, the decisions it makes, the risk category). It never overwrites an answer you already gave.
Related free tools
Package vulnerability database
Look up known vulnerabilities in npm, PyPI, Go, Cargo, RubyGems and Packagist packages, version by version, with the fix. Free, no account.
Dependency licence check
Upload a lockfile and see every dependency's licence. GPL, AGPL, unknown and changed licences are flagged for your kind of project. Free.
SBOM summary generator
Upload a lockfile and get an SBOM summary by email: components, licences, vulnerabilities and AI components. CycloneDX and SPDX files with Pro.
Last updated .