TOPIC 01 OF 06
Classify before you conclude
Risk tier first, verdict laterFaced with 'is this AI use allowed?', the instinct is to hunt for a rule. The more useful first move — the one the EU AI Act is built around — is to classify the use by how much risk it poses to people: roughly prohibited, high-risk, limited (transparency duties), or minimal. That triage tells you how much scrutiny the use needs and is the input a qualified reviewer turns into an actual determination.
A website FAQ chatbot and a tool that scores job applicants are both 'AI,' but they're worlds apart on risk: one mainly needs to tell users they're talking to a machine, the other decides people's access to work and carries heavy duties. Classifying first keeps you from treating them the same.
Risk tracks impact on people's rights and safety, not how advanced the technology feels. A simple tool making consequential decisions about people outranks a flashy one that suggests product names. This studio is training, not legal advice — classification guides scrutiny; it isn't a ruling.
Take one AI use you know and place it on the rough map — prohibited, high-risk, limited, or minimal — and write one line on why.