What does this EU AI Act checker tell me?

It places your AI system or general-purpose AI model into one of the AI Act's risk tiers — prohibited, high-risk, limited (transparency), or minimal — or, for GPAI models, whether systemic-risk obligations apply. It walks the Article 2, 5, 6, 50 and 51 logic to give a structured starting point.

Is the EU AI Act a directive my country still has to transpose?

No. The AI Act is a Regulation, directly applicable across the EU without national transposition. It applies in phases: prohibited practices from 2 February 2025, GPAI Chapter V obligations from 2 August 2025, and Article 50 transparency plus AI Office and national enforcement powers from 2 August 2026. Under Regulation (EU) 2026/1744 (Digital Omnibus on AI), high-risk Chapter III Sections 1-3 apply from 2 December 2027 (Annex III / Art. 6(2)) and 2 August 2028 (Annex I / Art. 6(1)). For a dated map of the 2 August 2026 hinge, see the guide at /learn/eu-ai-act-what-applies-from-2-august-2026.

We use open-source AI — are we exempt?

Only partly. Free and open-source AI is relieved of some obligations, but never from the Article 5 prohibitions, and not where the system is high-risk or falls under Article 50. Open-source GPAI without systemic risk gets a lighter transparency regime, not a full exemption.

My Annex III system feels low-risk — is it still high-risk?

Possibly not. Article 6(3) lets an Annex III system out of the high-risk tier if it only performs a narrow procedural task, improves a prior human activity, detects deviations without replacing human review, or is preparatory — but never if it profiles natural persons. This is a documented, fact-specific assessment.

Is this legal advice?

No. It is a free, structured starting point based on the Regulation's text and your inputs. Classification turns on the system's intended purpose and your operator role, and several boundaries are fact-specific or decided by the authorities. Confirm with your competent authority or counsel.

Do you store my answers?

No. The classification runs entirely in your browser. There is no form gate and we do not capture or store your inputs.

Free tool

EU AI Act risk-tier checker

Find out which EU AI Act risk tier your AI system or general-purpose AI model falls into — prohibited, high-risk, transparency, minimal, or GPAI — in about two minutes.

Classification follows Regulation (EU) 2024/1689 (AI Act) as amended by Regulation (EU) 2026/1744: Art. 2, 5, 6 + Annexes I & III, 50 and 51. It is a Regulation (no national transposition) and applies in phases: prohibited practices from 2 Feb 2025, GPAI Chapter V from 2 Aug 2025, Article 50 transparency and broader enforcement powers from 2 Aug 2026. High-risk Chapter III Sections 1-3 (except Art. 6(5)): 2 Dec 2027 (Annex III / Art. 6(2)) and 2 Aug 2028 (Art. 6(1) product route; Annex I Section B largely follows sectoral law under amended Art. 2(2)). A free, open-source licence does not exempt prohibited, high-risk or Art. 50 systems. This is a structured starting point, not legal advice.

What the EU AI Act requires, by risk tier

Obligations under the EU AI Act depend on the risk tier of your system and on your role in the value chain. Prohibited practices are banned outright, and minimal-risk uses carry no mandatory tier obligations; the substantive duties listed below fall on high-risk systems, general-purpose AI models, and limited-risk uses subject to transparency. Use the checker above to see which tier your use case most likely falls into.

•High-risk: risk-management system, data governance, technical documentation, logging, transparency to deployers, human oversight, accuracy/robustness/cybersecurity, conformity assessment + CE marking, registration (Art. 8–49).
•GPAI: technical documentation, information to downstream providers, a copyright policy, and a training-content summary (Art. 53) — plus, with systemic risk, model evaluation, systemic-risk mitigation, serious-incident reporting and cybersecurity (Art. 55).
•Limited risk: Article 50 transparency — disclose AI interaction; mark synthetic/deepfake content; inform people subject to emotion recognition or biometric categorisation.
•Obligations also depend on your role — provider, deployer, importer or distributor — which this tool does not determine.

Official sources

Jurisdiction: EU. Instrument: Regulation (EU) 2024/1689 (AI Act), applying in phases from 2 February 2025 (with GPAI enforcement and Article 50 from 2 August 2026; high-risk dates subject to Digital Omnibus amendments), as in force on the dates below.

Regulation (EU) 2024/1689 (AI Act), EUR-Lex (Last verified 2026-08-09)
European Commission: AI Act policy page (Last verified 2026-08-09)

FAQ

What does this EU AI Act checker tell me?: It places your AI system or general-purpose AI model into one of the AI Act's risk tiers — prohibited, high-risk, limited (transparency), or minimal — or, for GPAI models, whether systemic-risk obligations apply. It walks the Article 2, 5, 6, 50 and 51 logic to give a structured starting point.
Is the EU AI Act a directive my country still has to transpose?: No. The AI Act is a Regulation, directly applicable across the EU without national transposition. It applies in phases: prohibited practices from 2 February 2025, GPAI Chapter V obligations from 2 August 2025, and Article 50 transparency plus AI Office and national enforcement powers from 2 August 2026. Under Regulation (EU) 2026/1744 (Digital Omnibus on AI), high-risk Chapter III Sections 1-3 apply from 2 December 2027 (Annex III / Art. 6(2)) and 2 August 2028 (Annex I / Art. 6(1)). For a dated map of the 2 August 2026 hinge, see the guide at /learn/eu-ai-act-what-applies-from-2-august-2026.
We use open-source AI — are we exempt?: Only partly. Free and open-source AI is relieved of some obligations, but never from the Article 5 prohibitions, and not where the system is high-risk or falls under Article 50. Open-source GPAI without systemic risk gets a lighter transparency regime, not a full exemption.
My Annex III system feels low-risk — is it still high-risk?: Possibly not. Article 6(3) lets an Annex III system out of the high-risk tier if it only performs a narrow procedural task, improves a prior human activity, detects deviations without replacing human review, or is preparatory — but never if it profiles natural persons. This is a documented, fact-specific assessment.
Is this legal advice?: No. It is a free, structured starting point based on the Regulation's text and your inputs. Classification turns on the system's intended purpose and your operator role, and several boundaries are fact-specific or decided by the authorities. Confirm with your competent authority or counsel.
Do you store my answers?: No. The classification runs entirely in your browser. There is no form gate and we do not capture or store your inputs.

By ISMS Copilot. Classification follows Regulation (EU) 2024/1689 (AI Act) as amended by Regulation (EU) 2026/1744: Art. 2, 5, 6 + Annexes I & III, 50 and 51. It is a Regulation (no national transposition) and applies in phases: prohibited practices from 2 Feb 2025, GPAI Chapter V from 2 Aug 2025, Article 50 transparency and broader enforcement powers from 2 Aug 2026. High-risk Chapter III Sections 1-3 (except Art. 6(5)): 2 Dec 2027 (Annex III / Art. 6(2)) and 2 Aug 2028 (Art. 6(1) product route; Annex I Section B largely follows sectoral law under amended Art. 2(2)). A free, open-source licence does not exempt prohibited, high-risk or Art. 50 systems. This is a structured starting point, not legal advice.

Browse all 21 free compliance tools

Ready to do compliance work faster?

Built for speed, accuracy, and audit-ready output.

Try ISMS Copilot