Where does your AI land under the EU AI Act?

Where does your AI land under the EU AI Act?

Six questions about one AI use case. Out comes the risk class, the obligations that apply to your role, and the deadlines. Plain language instead of legalese, results in English or German.

1 · Your role

We use AI that we bought or licensed (deployer) We build or sell the AI system (provider)

2 · Does the system do any of these?

The banned list. Leave everything unchecked if none apply.

Manipulates or deceives people into harmful behavior Exploits age, disability, or economic hardship Social scoring with consequences in unrelated contexts Predicts who will commit crimes, based on profiling Scrapes faces from the internet into recognition databases Reads the emotions of employees or students Infers sensitive traits (origin, politics, religion, orientation) from biometrics Live facial recognition in public spaces for policing

3 · Where does it make or shape decisions?

The high-risk areas. Leave everything unchecked if none apply.

Hires, screens, promotes, monitors, or assigns work to people Decides on credit, insurance pricing, public benefits, or emergency calls Admits, grades, or proctors students Identifies people by biometrics Runs safety functions in critical infrastructure Supports police work Migration, asylum, or border decisions Supports courts or elections Is a safety component of a regulated product (machinery, medical devices, toys, vehicles)

5 · Does it talk, generate, or read people?

The transparency triggers. Leave everything unchecked if none apply.

People chat or talk with it (chatbot, voice bot) It generates text, images, audio, or video It runs permitted emotion recognition or biometric categorisation

6 · Result language

One use case per check. Nothing stored. Orientation, not legal advice.