Topic taxonomy
Each event carries one or more topics from this fixed list. Use them in subscription filters and API queries.
| Topic | Covers |
|---|---|
gpai | General-Purpose AI obligations (Art. 51-55, Code of Practice) |
high-risk | High-risk AI systems under Annex III (Art. 6 et seq.) |
prohibited | Prohibited practices (Art. 5) |
transparency | Transparency obligations (Art. 50, Art. 13) |
data-governance | Data quality and governance (Art. 10) |
human-oversight | Human oversight obligations (Art. 14) |
robustness | Accuracy, robustness (Art. 15) |
cybersecurity | Cybersecurity for AI systems (Art. 15(5)) |
fundamental-rights | Fundamental Rights Impact Assessment (Art. 27) |
standards | Harmonised standards (CEN/CENELEC, ISO, ETSI) |
enforcement | Enforcement, penalties, supervisory actions, AI Board opinions |
sandbox | Regulatory sandboxes (Art. 57) |
national-de | National implementation in Germany (BNetzA, BfDI, BSI, etc.) |
national-fr | National implementation in France (CNIL, DGE, etc.) |
national-nl | National implementation in the Netherlands (RDI, AP, etc.) |
national-other | Other Member States (ES, IT, IE, SE, etc.) |
meta | Process / organisational / EU AI Office structural updates |
Multi-topic events
Most non-trivial events carry 2-4 topics. A typical GPAI guidance update
might be tagged ["gpai", "transparency"]; an Annex III
amendment might be ["high-risk", "fundamental-rights"]; a
national sandbox launch is usually ["national-de", "sandbox"].
Tagging is LLM-driven
Tags are extracted by our enrichment model from the source's title, body, and any explicit Article references. The model is constrained to this exact list — no free-form tags.
Subscription matching
An event matches your subscription if at least one of its topics
appears in your subscription's topics array. Plus optional
source_filter, severity_min.