Skip to content
Klarvo

EU AI Act guide

One law. Four tiers. Find yours.

Regulation 2024/1689 is the world's first comprehensive AI law. It sorts every AI system into one of four tiers by the risk it poses — and attaches different rules to each. Here's the whole taxonomy, in plain English.

Banned outright

Prohibited

Article 5

Eight AI practices are simply illegal in the EU. No risk-management plan rescues them.

Read more →

Typical examples

  • Social scoring
  • Workplace emotion recognition
  • Untargeted face scraping
  • Manipulative dark-pattern AI

When it bites

Enforceable since 2 February 2025

Maximum penalty

€35M or 7% of global turnover

Heavy obligations

High-risk

Annex III

AI used in eight sensitive contexts. Conformity assessment, human oversight, logging, monitoring.

Read more →

Typical examples

  • Recruitment & CV screening
  • Credit scoring
  • Medical devices
  • Biometric identification

When it bites

From 2 August 2026 (Omnibus may delay to Dec 2027)

Maximum penalty

€15M or 3% of global turnover

Transparency

Limited risk

Article 50

AI that interacts with people or generates content must disclose itself. Where most SMEs land.

Read more →

Typical examples

  • Chatbots
  • AI-generated content
  • Deepfakes
  • Biometric inference

When it bites

From 2 August 2026 — not moving

Maximum penalty

€15M or 3% of global turnover

No AI-Act duties

Minimal risk

Article 4 only

The vast majority of AI. No AI-Act-specific obligations beyond staff literacy. GDPR still applies.

Read more →

Typical examples

  • Spam filters
  • AI search ranking
  • AI photo cleanup
  • Analytics

When it bites

AI literacy applies since 2 February 2025

Maximum penalty

Explainer hub

Read it once. Then let Klarvo hold the implementation.

The deadline doesn't move.

Article 50 transparency obligations apply on 2 August 2026. Klarvo's free tier classifies your first system and gives you the obligation list in minutes.

Free tier · Full KlarvoEngine classification · No credit card