EU AI Act
Global founders selling into the EU need clarity fast. Get your risk category and next compliance steps without guesswork.
Informational guidance only. Not legal advice.
EU AI Act risk classification is the process of determining whether an AI system is prohibited, high risk, limited risk, or minimal risk based on how it is used and who it affects.
For builders, this matters because risk level directly changes compliance work: documentation, transparency, human oversight, post-market monitoring, and go-to-market requirements.
A quick classification early in product design helps teams avoid costly rework and ship responsibly in the EU market.
Use this side-by-side view to compare category definitions, example systems, and likely compliance expectations.
| Category | Definition | Example systems | Compliance burden | Likely obligations |
|---|---|---|---|---|
| Prohibited Risk | AI uses that are considered unacceptable in the EU because they create clear harm to rights, safety, or democratic values. | Social scoring, manipulative systems exploiting vulnerabilities, certain real-time remote biometric identification use cases. | Not allowed in the EU market. | No route to compliance for prohibited practices; redesign is required. |
| High Risk | AI systems used in sensitive contexts where outcomes can significantly affect people. | Hiring and worker management tools, education assessment, access to essential services, healthcare or critical infrastructure support systems. | High compliance burden before and during deployment. | Likely requires technical documentation, risk management, human oversight, monitoring, and post-market controls. |
| Limited Risk | AI systems that are generally allowed but require transparency toward users. | Chatbots, synthetic media generation, AI assistants where users should know they are interacting with AI. | Moderate compliance burden focused on transparency. | Likely requires AI disclosure, labeling of generated content, and clear user-facing notices. |
| Minimal Risk | AI systems with low impact on rights and safety in typical product usage. | Spam filtering, basic recommendation support, low-impact productivity features. | Low mandatory burden under AI Act risk framework. | Usually no specific AI Act duties, but general obligations (for example GDPR) can still apply. |
Start with this lightweight product-style demo. It gives a preliminary classification before running a full scan.
Answer a few quick questions to get a preliminary AI Act risk classification.
Used on the public report page and social preview card.
Progress: 0/5 questions answered.
Complete all questions to unlock a reliable preliminary classification.
AI Act risk classification is the EU framework that groups AI systems into prohibited, high-risk, limited-risk, or minimal-risk categories based on impact and context of use.
High-risk AI typically includes systems used in areas like hiring, education, law enforcement, critical infrastructure, and healthcare where decisions can materially affect people.
Most chatbots are usually treated as limited risk and must meet transparency duties, such as informing users they are interacting with AI. Risk can increase depending on use case.
Prohibited AI practices are uses considered unacceptable under the EU AI Act, such as certain manipulative or exploitative systems and specific biometric surveillance use cases.
Yes. Startup status does not remove AI Act obligations. If your product is offered in the EU or affects people in the EU, you should assess classification and related duties.
Yes. Even if your AI is likely minimal risk under the AI Act, you can still have obligations under laws such as GDPR, consumer protection, and sector-specific rules.
No. This page and tool provide informational guidance and a preliminary assessment only. They do not replace advice from qualified legal professionals.
Explore deeper compliance guidance and practical implementation playbooks.
Move from preliminary answers to a structured compliance report in minutes.