EU AI Act
Every obligation, with its status.
Each AI Act obligation we track is published on our trust board, marked implemented, partial or unverified, with the evidence behind it.
Most AI Act obligations now apply, including the Article 50 transparency duties.
What the law requires
EU AI Act overview
The EU AI Act (Regulation (EU) 2024/1689) sets duties by risk tier. Every organisation that uses AI at work must take measures to promote AI literacy among its staff (Art. 4), and Article 50 sets transparency duties for AI systems that talk to people or generate content.
Who it applies to
Providers that place AI systems on the EU market, wherever they are established, and organisations in the EU that use AI systems (Art. 2). Use by a person for a purely personal, non-professional activity is excluded (Art. 2(10)).
Risk-based classification
AI systems are classified by risk. A document search and chat assistant that is not marketed for an Annex III use is limited-risk, with the transparency duties of Article 50. Our written classification of Nordvec is still open on the trust board.
Penalties
Fines reach EUR 35 million or 7% of worldwide annual turnover for prohibited practices (Art. 99(3)), and EUR 15 million or 3% for most other obligations (Art. 99(4)). For small and medium-sized enterprises, the lower of the two amounts applies (Art. 99(6)).
Nordvec and the AI Act
Where Nordvec stands
As a limited-risk system, Nordvec owes the transparency duties of Article 50. The logging and oversight duties of Articles 12 and 14 bind high-risk systems only. Each obligation's status is on the trust board.
Article 12, high-risk systems
Audit trail
Article 12 logging binds high-risk systems, which Nordvec is not. Questions, document views and administrative actions are still recorded in an append-only audit trail, without the text of questions or answers.
Append-only and hash-chained.
Article 50(1)
The assistant says it is AI
The chat tells each person that they are talking to an AI assistant and that its answers may contain errors. Answers from your documents cite the passages they used.
Disclosure shown in the chat.
Article 14, high-risk systems
Who can review use
Human-oversight duties under Article 14 bind high-risk systems. In Nordvec, workspace roles decide who can view the audit trail, and a member with that permission can review and export it.
Audit trail access set by role.
Article 50(2)
An AI mark on exported files
Documents you export in Markdown, Word, plain text or rich text carry a machine-readable mark stating that AI generated them. A printed copy carries a visible footer that says the same.
Marked at export.
Enforcement timeline
Key enforcement dates
February 2, 2025
Prohibited AI practices
Bans on social scoring, manipulative techniques, and real-time remote biometric identification in public spaces for law enforcement, with narrow exceptions (Art. 5).
ActiveAugust 2, 2025
General-purpose AI rules
Documentation and copyright duties for providers of general-purpose AI models (Art. 53).
ActiveAugust 2, 2026
General application
Most remaining obligations apply, including the Article 50 transparency duties. Obligations for high-risk systems apply later.
Active
Free tools
Check your own Article 50 duties
Two free tools built on the same cited legal grounding as this page. Both run in your browser; the optional signed export sends only a hash of the policy.
Read the obligations first.
Each AI Act obligation is published with its status and the evidence behind it.