Art.113

Article 113 – Entry into Force and Application

General SME Relevant ~2 min read · 480 words

Article 113 defines when each part of the EU AI Act takes legal effect. Following the Digital Omnibus provisional agreement of 7 May 2026, the high-risk AI application dates have been amended. The revised timeline has five key dates — with the main high-risk Annex III deadline now set at 2 December 2027.

! High compliance impact for SMEs

WHAT THE ARTICLE IS ABOUT

The application timeline — updated by the Digital Omnibus on AI

Article 113 establishes the phased application schedule of the EU AI Act. The Digital Omnibus provisional agreement of 7 May 2026, pending formal adoption, has amended the high-risk AI application dates. All other dates remain unchanged. The regulation is binding and directly applicable in all member states — no national transposition required.

WHAT IT SAYS

Five application dates — two amended by the Digital Omnibus

1 August 2024 — Entry into force. The Act entered the EU legal order. No compliance obligations apply yet
2 February 2025 — Chapter I (General Provisions) and Chapter II (Prohibited AI Practices, Article 5) apply. The prohibition on unacceptable risk AI is live from this date. Applied as scheduled — not changed by the Omnibus
2 August 2025 — Chapter V (GPAI models), Chapter VII (Governance), Article 78 (Confidentiality) and most of Chapter XII (Penalties) apply. Applied as scheduled — not changed by the Omnibus
2 December 2026 — New prohibition on AI generating non-consensual intimate imagery and CSAM applies. Article 50(2) watermarking obligations apply. New date introduced by the Omnibus
2 December 2027 — Full high-risk AI obligations for Annex III standalone systems apply. This replaces the original 2 August 2026 date. Amended by the Digital Omnibus
2 August 2028 — Full high-risk AI obligations for Annex I product-embedded systems apply. This replaces the original 2 August 2027 date. Amended by the Digital Omnibus
2 August 2030 — High-risk AI systems used by public authorities must comply by this date

WHO IS AFFECTED

Every business operating AI systems in the EU — the dates determine your compliance deadline

All providers, deployers, importers and distributors of AI systems in the EU
The specific obligations and their application dates vary by system type — check which date governs your specific system against this updated timeline

WHAT IT MEANS FOR SMES

Map your obligations to the correct updated date and work backwards

Already in force and unchanged: Article 5 prohibited practices (February 2025), GPAI obligations (August 2025), governance provisions
New December 2026 date: watermarking of AI-generated content and the new nudifier/CSAM prohibition — if you operate generative AI systems, this is your next compliance milestone
December 2027 is your main deadline if you deploy standalone high-risk AI under Annex III — employment screening, credit scoring, biometrics, education AI, and similar; 16 months more than originally planned but the institutions have signalled no further extensions
August 2028 applies only to AI embedded in Annex I regulated products — medical devices, machinery, vehicles
The Omnibus is provisional — formal adoption is expected before 2 August 2026; until that happens, the original dates technically remain law; plan against the new dates but do not stop compliance preparation
Small mid-cap enterprises (under 750 employees, under €150 million turnover) now benefit from the same facilitated procedures as SMEs under the Omnibus amendments

Article 5 — Prohibited practices (applied February 2025, extended by Omnibus with new prohibition from December 2026)
Article 6 — High-risk classification (determines which systems face the December 2027 deadline)
Article 111 — Transitional arrangements (grace periods for already-deployed systems under the updated timeline)
Article 50 — Transparency obligations (watermarking deadline moved to December 2026)

← Back to the Guide