Determine your EU AI Act risk class

An initial classification in a few questions

Find out in a few questions whether your AI system is likely high-risk under the EU AI Act and which duties roughly apply, indicative, not legally binding.

The EU AI Act assigns every AI system to one of four risk classes. Prohibited practices are banned. High-risk systems carry the most demanding duties, for example where the AI is a regulated product or its safety component under Annex I, or is used in a sensitive area under Annex III. Limited-risk systems mainly face the transparency obligation, and minimal-risk systems are subject to hardly any specific requirements. Where your system lands turns on its purpose and its deployment context, not on the technology alone.

This self-check does not replace a legal review. It gives you a first, honest tendency and a starting point for the conversation. All answers stay in your browser; no input is transmitted.

Is your AI system likely high-risk?

Answer every question with yes or no. You receive an instant, indicative classification together with the next steps.

Is your AI system a regulated product or a safety component of one, for example a medical device, machinery or a vehicle, under Annex I? Yes No

Do you use AI in a high-risk area under Annex III, such as biometrics, critical infrastructure, education, employment and recruitment, essential private or public services, law enforcement, migration or the administration of justice? Yes No

Does the system's output influence significant decisions about natural persons, for example applicant selection, creditworthiness or access to benefits? Yes No

Is the system placed on the EU market or used for people in the EU? Yes No

Is there no continuous, effective human oversight that can review and override the output? Yes No

What comes after the self-check?

We turn the indicative tendency into a solid, documented classification, legally led and tailored to your concrete use case.

The AI Governance Check is the clean next step: we review your system's classification on a case-by-case basis, map the EU AI Act duties and deliver a prioritised, board-ready roadmap with traceable documentation. That way you know whether and when which duties apply.

If you need ongoing support rather than a one-off assessment, our AI Officer takes over continuous AI governance: maintaining your AI inventory, keeping policies and evidence current, and accompanying new models and use cases before they go live.

Both are legally led: where a SIDD lawyer advises in a legal capacity, your information may be covered by professional secrecy under Art. 321 of the Swiss Criminal Code. We clarify the exact scope per mandate.

Why SIDD for the EU AI Act classification

Classifying an AI system is first a legal question, backed by in-house technical capability. That is exactly where our focus lies.

Legally led classification

Whether a system is high-risk turns on the legal terms of the EU AI Act and its interplay with product-specific law. The classification is led by doctorate-level lawyers with CIPP/E, not just a tool.

Engineering and pen testing for AI

With three dedicated AI services, AI Governance Check, AI Officer and AI Security, we cover the EU AI Act from initial classification through ongoing governance to penetration testing of LLM and agent systems.

Multilingual and independent

We advise in German, French and English, relevant for French-speaking Switzerland and EU parent companies. Because we do not sell an in-house SOC, our recommendations stay independent.

Swiss professional secrecy

Where a SIDD lawyer advises in a legal capacity, your information may be covered by professional secrecy under Art. 321 of the Swiss Criminal Code, in addition to contractual confidentiality. We clarify the exact scope per mandate.

Frequently asked questions about the EU AI Act risk class

Which risk classes exist in the EU AI Act?

The EU AI Act has four tiers: prohibited practices, high-risk systems, limited-risk systems and minimal-risk systems. Prohibited practices are banned, high-risk systems carry the most demanding duties, limited risk mainly triggers the transparency obligation, and minimal risk involves hardly any specific requirements.

What does high-risk mean in practice?

High-risk means the system must meet the strictest duties: documented risk management, data governance, technical documentation and logging, transparent information, effective human oversight and a conformity assessment with a CE link. A system is typically high-risk where it is a regulated product or its safety component under Annex I, or is used in a sensitive area under Annex III.

From when do the duties for high-risk AI apply?

The application deadlines for high-risk AI fall from 2026/2027, depending on whether the system falls under Annex I or Annex III. These deadlines are, however, under revision within the Digital Omnibus and may shift. They should be checked case by case; we track the developments and map your specific timeline.

What applies to general-purpose AI models (GPAI)?

General-purpose AI models carry their own duties, such as technical documentation, information for downstream providers and respect for copyright, with additional, stricter requirements for models with systemic risk. These duties apply regardless of the risk class of your specific application and can apply on top where you use or adapt such a model.