Living document

EU AI Act regulatory changelog

The regulation has moved once already and will move again. Every change that affects how AI systems are classified, recorded here — with what it meant in practice.

Last updated 8 August 2026 · Regulation (EU) 2024/1689 as amended by the June 2026 Omnibus · Self-assessment, not legal advice

The EU AI Act is a moving target — it has already been amended once since adoption, and it will move again. This page records every change that affects how Reglynn classifies systems, most recent first. Each entry says what changed in the regulation, who it affects, and what changed in our methodology as a result.

If you hold a Reglynn report: material changes here trigger your 12-month re-issue — reply to your report email and we re-run your assessment against the current rules, once, free.

2 August 2026 — enforcement begins

What changed: National authorities gained their fining powers. Article 5 (prohibited practices) and Article 4 (AI literacy), in force since February 2025, became enforceable. Article 50 transparency duties — chatbot disclosure, synthetic-content labelling, deepfake disclosure — applied and became enforceable the same day. Article 53 GPAI obligations, applicable since August 2025, also became enforceable.

Who it affects: Everyone — but in practice, chatbot and generative-AI products most, because those are the systems whose duties were not deferred. High-risk timelines are unaffected (still December 2027 / August 2028).

Reglynn changes: Reports now flag Articles 4, 5, 50 and 53 as live rather than upcoming. No classification logic changed.

June 2026 — the Omnibus amendment

What changed: The most consequential amendment to date. Annex III high-risk obligations deferred from August 2026 to 2 December 2027; Annex I (AI in regulated products) to August 2028. Article 50 transparency was not deferred. Article 99(6) fine structure for SMEs confirmed: the applicable fine is the lower of the fixed amount and the percentage of worldwide turnover.

Who it affects: High-risk providers gained 16 months — which most need in full, since conformity assessment realistically takes 12–18 months. Chatbot and generative products gained nothing, and much published guidance failed to notice; a great deal of content still tells them they have time they don't have. Our deadline table carries only post-Omnibus dates.

Reglynn changes: Methodology v1.0 built directly on the amended text. All deadline outputs use post-Omnibus dates; the report carries a regulation-version stamp naming the June 2026 amendment; the Article 99(6) rule was hard-coded into the fine framing so no SME is ever quoted the €15M headline number.

August 2025 — GPAI obligations apply

What changed: Article 53 obligations for general-purpose AI model providers took effect: technical documentation (Annex XI), information for downstream providers (Annex XII), a copyright and text-and-data-mining policy, and a public training-data summary. Enforcement followed on 2 August 2026.

Who it affects: Foundation-model providers directly. The trap is downstream: teams that substantially retrained a model — including serious fine-tunes done in 2024–25 for performance reasons — may hold provider duties without knowing. Light fine-tuning and API use stay downstream. The threshold.

February 2025 — first provisions in force

What changed: Article 5 prohibitions (social scoring, real-time public biometric identification, workplace and classroom emotion recognition, untargeted facial-image scraping, exploitative manipulation) and Article 4 AI literacy came into force, with enforcement deferred to August 2026.

1 August 2024 — entry into force

What changed: Regulation (EU) 2024/1689 entered into force with a staggered application calendar. Everything above happened on that calendar — as amended by the Omnibus.

What we're watching

When any of these lands, it gets an entry here, a methodology version bump where it changes classification, and — where material — a re-issue notice to report holders.

Check where you stand under the current rules

Two minutes against the post-Omnibus text — not last year's dates.

Start the free scan

No card, no account. Full gap report from €59. See a real report first.

Self-assessment, not legal advice. This page and the Reglynn report are a structured self-assessment based on the EU AI Act (Regulation 2024/1689, as amended by the June 2026 Omnibus). They indicate where you likely stand and what to verify with a qualified advisor. Reglynn is not a notified body and issues no certification.

How this was made. Reglynn reports are generated by an AI system and reviewed before delivery. Reglynn is a Limited Risk AI system under Article 50; we disclose AI involvement wherever it occurs, including in our chat assistant. Our own self-assessment.