The law doesn't ask whether you use AI.
It asks whether you can prove it.
As of August 2, 2026, the regulation applies throughout the European Union. This page summarizes its requirements and when they apply, and allows you to classify your own system using the same eleven questions that the platform uses. No registration or data submission required.
Written for senior management and the legal department. Free of unnecessary technical jargon.
Deadline for high-risk systems listed in Annex III, following the extension provided for in the AI Omnibus (effective July 27, 2026).
or 7% of worldwide annual turnover, for prohibited practices.
or 3% of worldwide annual turnover, for failing to comply with high-risk obligations.
These obligations apply to whoever deploys the system, not just whoever develops it.
Dates under Regulation (EU) 2024/1689, as amended by the AI Omnibus. Source: European Commission.
The regulation does not regulate “AI” per se. It regulates specific uses based on the harm they may cause.
The more a system affects people's lives, the more the law requires. Most business applications involve minimal or limited risk. Decisions regarding credit, employment, insurance, or public services are high-risk, and that is where evidence is needed.
Prohibited practices under Article 5. These may not be marketed or used in the EU.
Annex III and safety components. Risk management, documentation, record-keeping, human supervision, and conformity assessment.
Chatbots, generated content, emotion recognition. Transparency requirements under Article 50.
The rest. No specific obligations beyond AI literacy and voluntary best practices.
What level is
's system?
Eleven questions. The risk level, reasons and obligations update as you answer. It follows the same logic as the initial assessment at V-PROOF; the result is for guidance only and does not replace a formal assessment.
Subliminal manipulation, exploitation of vulnerabilities, social scoring, predictive policing based solely on profiling, emotion recognition in the workplace or in educational institutions, biometric categorization of sensitive attributes, and real-time remote biometric identification in public spaces. Starting in December 2026, this will also include the generation of non-consensual sexual or intimate content.
Machinery, medical devices, vehicles, toys, elevators…
Hiring, credit, benefits, medical triage, qualifications, access to services.
A chatbot, an assistant.
- They will appear here as you answer.
- They will appear here as you answer.
For reference only. None of the information you enter leaves your browser. The formal classification is determined by the platform’s initial evaluation and is recorded with an attestation.
Each article makes a specific request. Each one leaves specific evidence.
For a high-risk system, these are the main obligations and the proof that V-PROOF generates for each one. What is recorded is the cryptographic fingerprint of the event, never the data itself.
A general overview of the regulations. This does not constitute legal advice; specific application depends on each system and its formal evaluation.
An assistant that screens CVs. How it is governed, step by step.
Learn how an organization registers the system, classifies it as high risk (Annex III, employment), obtains approval with human oversight, and and produces its first third-party-verifiable evidence.