News follow-up · Regulation
The EU AI Act: what happened since
The Act took effect in August 2024 and has been applied in stages. In July 2026 an amending regulation pushed the high-risk rules back by up to 16 months. The dates, from the Commission's own consolidated text.
What this site reported
On 10 June 2024 the earlier version of this site reported that the European Parliament had approved the first general law on artificial intelligence, with rules sorted by the risk of each use.
What the record shows now
The Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024.[1] The Commission's note of that day described the four tiers it set: minimal risk with no obligations, transparency duties for systems such as chatbots, strict requirements for high-risk systems such as medical software or recruitment tools, and bans on uses such as social scoring by governments or companies.[1] The rules for general-purpose AI models were to apply twelve months later, with the Commission's AI Office supervising them.[1]
The timetable then changed. On 16 June 2026 the European Parliament adopted its first-reading position on the Commission's proposal COM(2025)0836, the "Digital Omnibus on AI", and the adopted text records that the position corresponds to the final act, Regulation (EU) 2026/1744.[3] The Commission's consolidated text of the AI Act, as at 27 July 2026, now reads as follows at Article 113:[2]
| Part of the Act | Original date | Date after the Digital Omnibus on AI |
|---|---|---|
| Chapters I and II (general provisions, prohibited practices) | 2 February 2025 | unchanged, except new provisions added to Article 5, which apply from 2 December 2026 |
| General-purpose AI models, governance, penalties (Chapters V, VII, XII and others) | 2 August 2025 | unchanged |
| Most of the rest of the Act | 2 August 2026 | unchanged |
| Articles 102 to 110 (amendments to other EU laws) | 2 August 2026 | 27 July 2026 |
| High-risk systems listed in Annex III | 2 August 2026 | 2 December 2027 |
| High-risk systems that are, or sit in, products under Annex I | 2 August 2027 | 2 August 2028 |
So the obligations for the high-risk uses that matter most to public bodies, the Annex III list, which covers areas the Commission's own FAQ names as employment and law enforcement, now start on 2 December 2027 instead of 2 August 2026.[2][4] The FAQ describes the mechanism: the high-risk rules were tied to the availability of harmonised standards and other support tools, with a backstop of at most 16 months of delay for Annex III and 12 months for Annex I.[4] The amended article now states the backstop dates outright.[2]
Two things did not move. The prohibitions and the AI literacy duty have applied since 2 February 2025, and the general-purpose model rules since 2 August 2025.[2] For the model rules, the FAQ says the AI Office's enforcement powers were in place for 2 August 2026, including fines of up to 3% of global annual turnover and the power to require a model to be withdrawn from the market.[4]
What it means for public bodies
A public body in the EU that uses, or buys, an AI system in an Annex III area has until 2 December 2027 before the provider and deployer obligations of Chapter III apply to it, and the supervision of EU institutions themselves sits with the European Data Protection Supervisor.[2][4] The transparency duties and the bans are already in force.[2] Anyone reading a 2024 summary of the Act, including the one this site ran, should replace its dates with the table above.
See also: AI in the public sector and Generative AI in government for the United States rules.
Sources
- European Commission, news article, AI Act enters into force, 1 August 2024. Read 29 October 2026.
- European Commission, AI Act Service Desk, Article 113: Entry into force and application, text based on the EUR-Lex consolidated version of Regulation (EU) 2024/1689 as at 27 July 2026, with the Digital Omnibus on AI amendments marked. Read 29 October 2026.
- European Parliament, texts adopted, P10_TA(2026)0198, Simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI), legislative resolution of 16 June 2026 on proposal COM(2025)0836. Read 29 October 2026.
- European Commission, AI Act Service Desk, Frequently asked questions, sections Digital Omnibus and Governance and Enforcement. Read 29 October 2026.
How this page was checked: the application dates were read from the Commission's consolidated Article 113 text and compared with the FAQ; the amending regulation's number and the Parliament's vote date were read from the Parliament's adopted text. The four documents were read on 29 October 2026.