AI risk & governance
The EU law that sorts AI systems by how much harm they could cause and sets stricter rules the higher the risk.
Formal
An EU regulation, in force from 1 August 2024 and applying directly in every member state, that bans some uses of AI, sets duties for high-risk systems, requires openness about chat assistants and deepfakes, and places duties on makers of general-purpose AI models.
In plain English
Like food safety rules that barely touch a bakery selling bread but put a baby-food factory under close inspection - the rules grow with what could go wrong.
In practice
A Danish recruitment firm finds its CV-screening tool counts as high-risk, so before 2 December 2027 it must document the system, keep logs, test for unfairness and make sure a person can override it.
Why it matters
As the first broad AI law, it reaches organisations that build, sell or use AI in the EU, in stages - bans from February 2025, general-purpose model rules from August 2025, openness duties from August 2026, most high-risk rules from December 2027 (August 2028 for AI in products). Fines reach 7% of global turnover.