Article 3 of the AI Act defines 68 terms verbatim. If your team uses “AI system,” “general-purpose AI model,” or “deployer” inconsistently across documentation — that’s a compliance gap waiting to be found.
Answer: Article 3 of Regulation (EU) 2024/1689 defines 68 legal terms. In 2026, teams should prioritize Article 50 transparency and GPAI enforcement readiness. Under the 7 May 2026 AI Omnibus political agreement, high-risk rules move to 2 December 2027 for stand-alone systems and 2 August 2028 for product-embedded systems; verify final Official Journal text before relying on those dates. Last verified: 2026-06-22.
Article 3 of the EU AI Act defines 68 terms. Your legal team, product team, DPO, and documentation team all need to use these terms consistently and correctly across all documentation. Terminology inconsistency is exactly the kind of control weakness that auditors flag. Here's what's at stake:
When one team writes “AI provider” and another writes “AI developer” — they mean the same thing but an auditor sees inconsistency. When your risk assessment says “high-risk AI” but your technical documentation says “high-risk AI system” — that’s a documentation gap.
Sources checked 2026-06-22: Regulation (EU) 2024/1689, European Commission AI Act page, Council AI Omnibus political agreement, and AI Act Service Desk Article 50.
The CSV template contains every numbered Article 3 definition from Regulation (EU) 2024/1689, with source notes and practical categories for Confluence review. The visible list below is a preview, not the full legal table:
All 68 Article 3 definitions, verbatim from Regulation (EU) 2024/1689, pre-categorized. Submit your email and the packet is delivered to your inbox.
| AI Act Requirement | How We Help |
|---|---|
| Consistent terminology across all documentation | Compliance Scanner checks every Confluence page for deprecated or unapproved terms |
| Audit trail for terminology decisions | Full version history with who changed what, when, and why |
| Approval process for definition changes | Four-eyes approval workflow: draft → review → approved → deprecated |
| Evidence for regulators and auditors | One-click CSV export with complete approval chain and version history |
| Cross-team alignment (legal, product, DPO) | Single source of truth in Confluence where teams already work |
| Ongoing compliance as definitions evolve | Auto-scan on page changes catches new inconsistencies immediately |
See the full list of audit control weaknesses we address, mapped to real regulatory standards. For setup instructions and feature details, see the documentation. We also publish our transparent list of app limitations.
The downloadable CSV contains 68 Article 3 definitions plus a header row. The preview below is deliberately high-level so this page does not hand-maintain legal numbering that can drift from the template.
| # | CSV field | How it is used in Confluence |
|---|---|---|
| 1 | termName | The approved term teams should use in pages, policies, and templates. |
| 2 | definition | The source-backed Article 3 definition text to review before approval. |
| 3 | category | A practical grouping for navigation, ownership, and review workflow. |
| 4 | synonyms | Deprecated or informal alternatives the scanner can flag for review. |
| 5 | notes | The Article 3 source note that should be checked against the current legal text. |
Source of truth for exact rows: the emailed AI Act glossary packet. Legal source: Regulation (EU) 2024/1689. Regenerate the packet and table together if Article 3 is amended.
Article 3 of Regulation (EU) 2024/1689 contains 68 numbered legal definitions, ranging from “AI system” (Article 3(1)) to “downstream provider” (Article 3(68)). All 68 are defined verbatim in the regulation text.
Under the 7 May 2026 AI Omnibus political agreement, high-risk rules are planned for 2 December 2027 for stand-alone Annex III systems and 2 August 2028 for high-risk AI systems embedded in regulated products. Treat those as current planning dates and re-check the final Official Journal text before making a legal deadline call.
Fines under Article 99(4) cap at €15 million or 3% of global annual turnover for breaches of high-risk obligations. Prohibited AI practices under Article 5 carry a higher cap of €35 million or 7% of global turnover (Article 99(3)), which has applied since 2 February 2025. The penalty framework itself only became enforceable on 2 August 2025 (Article 113(b)).
Under Article 3(3), a “provider” develops an AI system or has one developed and places it on the market or puts it into service under its own name or trademark. Under Article 3(4), a “deployer” is any natural or legal person using an AI system under its authority, except in the course of a personal non-professional activity. A company buying an HR screening AI is typically a deployer; the vendor building it is the provider.
Annex III lists use cases that can make an AI system high-risk under Article 6(2): biometrics, critical infrastructure, education and vocational training, employment and worker management, access to essential services, law enforcement, migration/asylum/border control, and administration of justice and democratic processes. The AI Omnibus political agreement moves the stand-alone Annex III high-risk planning date to 2 December 2027, pending final text.
The AI Act is not coming — it’s here. First obligations are already in effect. Import the glossary template, assign your team to review definitions, and start scanning your documentation today.
Evaluate in Confluence Get Free TemplateCompliance for Confluence — approved terms, page scanning, and audit evidence for regulated teams in Confluence
NIS2 Directive Terminology — EU cybersecurity compliance terminology guide
DORA Regulation Terminology — financial services digital resilience terminology guide
FDA Terminology Management — 43 terms for pharma & medtech, 21 CFR Part 11 aligned
SOC 2 Terminology Management — 40 Trust Services Criteria terms for InfoSec & GRC teams
Compliance Guide — what auditors check and how we help
Glossary App Comparison — how Confluence glossary apps compare