01 /
Prohibited
Examples
Social scoring, biometric surveillance, manipulative AI, emotion recognition in the workplace and education
What you must do
Cannot be used at all. Banned since February 2025.
Regulation (EU) 2024/1689 · deadline 2 August 2026
EU AI Act,
explained in practice
If you use AI to make decisions about people, such as recruitment, credit, legal advice or healthcare, you are in scope. Here is what that means, when it hits, and where you start.
§ 01
Timeline
Five deadlines
Three have already passed. The next one, 2 August 2026, is the one that hits most mid-market firms with high-risk AI systems.
- 1 August 2024EU AI Act entered into force
- 2 February 2025Prohibited practices (Art. 5) apply
- 2 August 2025General-purpose AI models (Art. 51-55) apply
- 2 August 2026High-risk AI systems (Annex III) apply in fullNext
- 2 August 2027Remaining rules apply in full
§ 02
Classification
Four risk categories
02 /
High-risk
Examples
Recruitment, credit scoring, legal advice, healthcare, insurance, education, critical infrastructure
What you must do
Risk management, documentation, registration in the EU database, human oversight, transparency, post-market monitoring.
03 /
Limited
Examples
Chatbots, deepfakes, generative AI with output directed at citizens
What you must do
Clear labelling that the user is interacting with AI. Generative output must be marked.
04 /
Minimal
Examples
AI for spam filtering, inventory management, internal productivity
What you must do
No specific requirements. Voluntary codes of conduct encouraged.
§ 03
Free resource
8-page practical checklist
For mid-market executives. Risk classification for your systems, what you must document, where to start, sector-specific add-ons for law, accounting, finance and IT services.
- →Visual overview of the 5 deadlines
- →4 risk categories with concrete examples
- →12-point checklist per risk category
- →3 first actions you can take this week
Note: the checklist PDF is currently produced in Danish. An English translation is in progress.
§ 04
Next steps
When the checklist is not enough
Our one-day EU AI Act Quick Check delivers a complete AI system inventory, risk classification with article reference, and a 90-day roadmap to the 2 August 2026 deadline. Built on 11+ years of GRC platform experience from Archer.