Regulatory changelog
Every date this platform applies, and where it comes from
Telling a compliance officer the wrong applicability date is the worst error a product like this can make. So there is one source for these dates inside E-ARI, every surface reads from it, and this page shows you that source. If we are wrong, you can see that we are wrong.
What the Digital Omnibus changed
Regulation (EU) 2026/1744 (Digital Omnibus on AI), amending 2024/1689 came into force on 27 July 2026. It deferred the high-risk obligations and added two prohibitions. It did not move the Article 4 literacy duty or the Article 50 transparency duties, which is the part most summaries get wrong — those still apply on their original dates.
Two further prohibitions: non-consensual intimate imagery, and child sexual abuse material
now 2 December 2026 · Art.5
Marking of synthetic content for systems already on the market before 2 August 2026
now 2 December 2026 · Art.50(2)
High-risk obligations for standalone Annex III systems
now 2 December 2027 · Art.6(2), Art.9–17, Art.26, Annex III
High-risk obligations for AI embedded in regulated products
now 2 August 2028 · Art.6(1), Annex I
The full application timeline
- 2 February 2025applies now
Prohibited practices banned
Art.5
- 2 February 2025applies now
AI literacy duty — applies to every organisation using AI, not only high-risk
Art.4
- 2 August 2025applies now
General-purpose AI model obligations
Art.53 · Art.55
- 2 August 2026applies now
Transparency duties for AI that interacts with people or generates content
Art.50
- 2 December 2026105 days away · moved by the Omnibus
Two further prohibitions: non-consensual intimate imagery, and child sexual abuse material
Art.5
- 2 December 2026105 days away · moved by the Omnibus
Marking of synthetic content for systems already on the market before 2 August 2026
Art.50(2)
- 2 December 2027470 days away · moved by the Omnibus
High-risk obligations for standalone Annex III systems
Art.6(2) · Art.9–17 · Art.26 · Annex III
- 2 August 2028714 days away · moved by the Omnibus
High-risk obligations for AI embedded in regulated products
Art.6(1) · Annex I
- 2 August 20301444 days away
Legacy public-authority systems brought into scope
Art.111
When each engine was last checked against the text
Regulatory content in this product cannot be added without recording what it was checked against. These are the modules that encode the law, what each one decides, and the date it was last verified.
| Module | Decides | Checked |
|---|---|---|
| controls-derive.ts | How an obligation is satisfied: by an uploaded document, or by the classification record. The Art.5 prohibitions are settled by the screening determination — a prohibition is not evidenced by a document saying you do not do the thing — and a system classified prohibited fails rather than passes. | 2026-08-18 |
| ai-act-timeline.ts | Application dates for each part of the Act | 2026-08-04 |
| ai-act-scope.ts | Safety-component test, Art. 6(1a) exclusions, small mid-cap thresholds | 2026-08-04 |
| compliance/ai-act-obligations.ts | Obligation catalogue behind the gap radar and coverage matrix | 2026-08-04 |
| ai-act-annex.ts | Annex III areas, all eight Art. 5 prohibition limbs plus the two added by Reg. 2026/1744, and Art. 50 triggers, as data | 2026-08-14 |
| ai-act-classify.ts | Deterministic risk-tier rules and the trace behind each determination | 2026-08-06 |
| compliance/classifier.ts | Rationale prompt and citation guard over the rule-decided tier | 2026-08-06 |
| regulatory-mapping.ts | Pillar-to-obligation mapping across the Act, NIST AI RMF and ISO 42001 | 2026-08-04 |
| training-modules.ts | Article 4 literacy curriculum, including the risk-tier explainer | 2026-08-04 |
| ai-tool-catalog.ts | Per-tool risk notes, including which tools carry the Art. 50 transparency duty | 2026-08-04 |
| progression.ts | Which obligations gate each stage of the compliance journey | 2026-08-04 |
| vendor-questionnaire.ts | Supplier questions mapped to provider obligations, incl. Art. 51 GPAI | 2026-08-04 |
| plain-summary.ts | The timing paragraph on every results page | 2026-08-04 |
| marketing-engine.ts | Ground-truth block handed to the language model | 2026-08-04 |
Oldest verification: 2026-08-04 (ai-act-timeline.ts).
Instruments relied on
- Regulation (EU) 2024/1689 (Artificial Intelligence Act)in force 1 August 2024
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), amending 2024/1689in force 27 July 2026
- Commission Recommendation (EU) 2025/1099 on the definition of small mid-cap enterprisesnot binding
- Draft Commission guidelines on the classification of high-risk AI systems (19 May 2026) — DRAFT, consultation closed 23 July 2026, not adoptednot binding
- NIST AI Risk Management Framework 1.0not binding
- ISO/IEC 42001:2023 — AI management systemsnot binding
This is a record of what the platform applies, not legal advice. Dates and scope depend on how your specific systems are classified. How classification works.