Did the EU AI Act's high-risk obligations take effect in August 2026 on schedule?
Published
Current View
As of 23 September 2026, no—according to the European Commission’s official explanatory page, the principal high-risk obligations did not become applicable without delay on 2 August 2026. It reports that Annex III high-risk rules were extended to 2 December 2027 and Annex I regulated-product high-risk rules to 2 August 2028. The Commission describes the AI Omnibus as having entered into force on 27 July 2026, not merely as a pending postponement proposal. The general application milestone nevertheless held: the Commission says the Act became applicable on 2 August 2026 with exceptions, with transparency rules applying in August 2026. Prohibited practices and AI-literacy obligations had already applied from 2 February 2025, and governance rules and general-purpose AI (GPAI) obligations from 2 August 2025, subject to transitions. The remaining dates are schedules, not guarantees against further amendment. Besides the revised high-risk dates, the Commission identifies a new intimate-content/CSAM prohibition applying in December 2026. The supplied secondary timeline reports additional transitions in 2027 and through December 2030, detailed below. This is a qualified official-summary finding, not a complete legal-text verification: the evidence bank does not contain the operative regulation or amending instrument. The amendment’s identifier, publication and exact clauses, the original allocation of obligations, and the full proposal history remain unverified. The reviewed pages cannot establish that no other timeline-changing measure or proposal exists.
Reasons
- The Commission reports an AI Omnibus legislative proposal adopted on 19 November 2025, political agreement on 7 May 2026, and entry into force on 27 July 2026. It gives revised application dates of 2 December 2027 for Annex III high-risk systems and 2 August 2028 for Annex I product-related systems. Formal legislative adoption and publication details are not independently verified here.
- Already applicable according to the Commission: prohibited practices and AI literacy from 2 February 2025; governance and GPAI obligations from 2 August 2025; and general application from 2 August 2026 with exceptions. Its transparency summary specifies August 2026.
- The Commission places the new prohibition concerning AI-generated non-consensual intimate content and CSAM in December 2026. The secondary timeline specifies 2 December 2026 and also lists that date for Article 50(2) compliance by providers of qualifying synthetic-content systems placed on the market before 2 August 2026; those detailed provisions require official-text confirmation.
- The secondary timeline lists 2 August 2027 for compliance by providers of GPAI models placed on the market before 2 August 2025 and for operational national regulatory sandboxes. These detailed milestones are not independently confirmed against official legal text in this investigation.
- The secondary timeline describes the postponed high-risk provisions as Chapter III, Sections 1–3, except Article 6(5), and describes significant-design-change conditions for qualifying pre-existing systems. Its account supports avoiding a blanket claim that every high-risk-related provision is postponed, but the exact scope remains legally unverified.
- The secondary timeline reports later compliance dates of 2 August 2030 for high-risk systems intended for public-authority use and 31 December 2030 for qualifying Annex X large-scale IT systems placed on the market or put into service before 2 August 2027. Their precise scope requires official-text verification; August 2028 cannot be treated as the end of every reported transition.
Counter-evidence
- Against an unchanged 2 August 2026 high-risk milestone, the Commission expressly reports extensions to December 2027 and August 2028. This is official explanatory evidence, rather than an independently inspected amending regulation.
- Against interpreting the postponement as Act-wide, the Commission retains general application on 2 August 2026, earlier prohibitions/AI-literacy and governance/GPAI phases, and August 2026 transparency rules.
- The Commission initially attributes the extensions to a political agreement but then expressly reports entry into force on 27 July 2026. Reading the Omnibus as merely pending would disregard that later statement; the underlying legal instrument is needed to verify its formal status and clauses.
- Against deferring every high-risk-related provision, the secondary timeline limits its description to specified Chapter III sections and expressly excepts Article 6(5). The precise legal boundary remains unresolved.
- Against treating August 2028 as the final transition, the secondary timeline lists public-authority and qualifying legacy large-scale IT-system deadlines in 2030. These require confirmation from official legal text.
Limitations
- The assessment uses the supplied page snapshots retrieved before the asOf date; their publication dates are unknown.
- The Commission page is treated as authoritative explanatory reporting, not as a substitute for the operative legal texts.
- Applicability is distinguished from universal compliance or enforcement readiness.
- Future scheduled dates are not assumed to be immune from amendment.
- What are the exact operative provisions of Regulation (EU) 2024/1689 and its timeline-changing amendment, including the original allocation of high-risk obligations to the 2026 and later milestones? Neither permitted page supplies the official legal text itself.
- What is the AI Omnibus amendment's official regulation number, Official Journal publication date, formal adoption dates, and complete entry-into-force and application clauses? The Commission summary reports entry into force but those legal details were not verified.
- Do official legal texts confirm the secondary timeline's detailed Chapter III scope and Article 6(5) exception, legacy-system significant-design-change conditions, Article 50(2) transition, sandbox date, and 2030 transitional deadlines?
- What were the exact dates and mechanisms in the originally proposed postponement and subsequent legislative revisions, as distinct from the outcome summarized by the Commission?
- Are there any additional adopted amendments, corrigenda, or pending timeline-changing proposals not described on the two permitted pages?
- Will the remaining scheduled milestones remain unchanged through their arrival? No future legislative outcome can be established from this snapshot.
What to watch next
- Obtain the Official Journal texts of Regulation (EU) 2024/1689 and the reported AI Omnibus amendment, including identifiers, publication dates, operative application clauses and transitional provisions.
- Verify the original high-risk milestones and the postponement proposal’s initial mechanism, revisions and formal legislative adoption history.
- Check official legislative records for further pending or adopted timeline changes; absence from these two pages is not evidence of absence.
- Confirm secondary-only details concerning Article 6(5), legacy-system design changes, Article 50(2), sandboxes and the 2030 transitions.
- Track the arrival of remaining application dates and any intervening adopted amendments.