The EU AI Act Timeline: Every Compliance Deadline Explained

This AI Act timeline maps every EU compliance deadline in one place. The EU AI Act — Regulation (EU) 2024/1689 — does not switch on all at once. Article 113 phases its obligations in over several years, so “the AI Act deadline” is really a series of dates, each binding a different set of rules. This hub lays out every milestone, what becomes enforceable on each date, and what you should have done before it. As of mid-2026, the next deadline that actually binds new obligations is the Article 50 transparency regime on 2 August 2026 — and unlike the high-risk rules, it has not been pushed back.

The complete AI Act timeline at a glance

The three “pending” rows reflect the Digital Omnibus on AI, explained in full below. The first three rows and the 2 August 2026 Article 50 date are already settled in the regulation as adopted.

Phase by phase: what each deadline requires

1 August 2024 — Entry into force

The AI Act was published in the Official Journal on 12 July 2024 and entered into force twenty days later. Entry into force starts the clock, but almost none of the substantive obligations applied on this date — they were deliberately staggered under Article 113 to give providers, deployers and national authorities time to prepare.

2 February 2025 — Prohibited practices and AI literacy

The first substantive milestone. Two things became applicable: the prohibited practices in Article 5 (including manipulative or deceptive techniques causing significant harm, exploitation of vulnerabilities, social scoring by public authorities, certain real-time remote biometric identification in public spaces, untargeted facial-image scraping, and emotion inference in workplaces and education) and the AI literacy duty in Article 4, which requires providers and deployers to ensure their staff have a sufficient level of AI literacy.

2 August 2025 — GPAI, governance and penalties

This date brought obligations for providers of general-purpose AI (GPAI) models, the EU’s governance architecture (the AI Office, the AI Board, notified bodies and national competent authorities), confidentiality rules, and crucially the penalties framework in Chapter XII. From this point the fine tiers in Article 99 are live — even though many of the obligations they enforce only become applicable later.

2 August 2026 — Article 50 transparency (the next live deadline)

This is the deadline that matters most for the broadest range of organisations, because Article 50 is not limited to high-risk systems — it applies to any AI used in the four situations it covers. For a full walkthrough see our complete guide to Article 50 transparency obligations. Its four obligations are:

• AI-interaction disclosure (Art. 50(1)): providers must ensure people are informed they are interacting with an AI system, unless that is obvious from the circumstances and context of use.
• Synthetic-content marking (Art. 50(2)): providers of generative AI must mark AI-generated audio, image, video and text in a machine-readable format, detectable as artificially generated or manipulated.
• Deepfake labelling (Art. 50(4)): deployers who generate or manipulate deepfake content must disclose that it is artificially generated or manipulated — see our detailed guide to the deepfake labelling rules under Article 50(4).
• Public-interest text labelling (Art. 50(4)): AI-generated text published to inform the public on matters of public interest must be disclosed as such, subject to the Article’s exceptions.

What to have done by 2 August 2026: inventory every AI system you provide or deploy that interacts with people or produces content; decide which of the four obligations applies to each; implement the disclosure and marking mechanisms (the Commission’s Article 50 guidelines of 8 May 2026 and the Code of Practice on Transparency of AI-Generated Content, published 10 June 2026, are your reference points); and document the basis for any “obvious from the circumstances” exception you intend to rely on.

2 December 2026 — the Article 50(2) grace period for legacy generative AI

One narrow piece of relief sits inside the Article 50 regime. Generative AI systems already on the market before 2 August 2026 get until 2 December 2026 to comply with the machine-readable marking duty in Article 50(2) only. Three points are easy to get wrong: it covers only Art. 50(2), not Art. 50(1), (3) or (4), which apply from 2 August 2026 with no transition; it benefits only legacy systems, so anything newly placed on the market from 2 August 2026 complies from day one; and it arrives through the Digital Omnibus, so it is itself pending formal adoption. Separately, content that was generated and made available before 2 August 2026 does not have to be marked or labelled retroactively.

2 December 2027 — high-risk stand-alone systems (Annex III)

Obligations for stand-alone high-risk systems listed in Annex III — used in areas such as employment, education, essential services, law enforcement and migration — were originally set for 2 August 2026. The Digital Omnibus defers them to 2 December 2027, a sixteen-month extension intended to give the market time to finish the harmonised standards and conformity-assessment tools high-risk compliance depends on.

2 August 2028 — high-risk systems embedded in regulated products (Annex I)

High-risk AI that is a safety component of products already governed by EU sectoral legislation (Annex I — for example radio equipment, lifts and medical devices) was originally due on 2 August 2027. The Omnibus moves it to 2 August 2028, aligning it more closely with the underlying product-safety frameworks.

The Digital Omnibus: why some deadlines moved (and why they’re not final yet)

The Digital Omnibus on AI is the first package of amendments to the AI Act since its adoption. The Commission tabled it on 19 November 2025; the Parliament and Council reached a provisional political agreement on 7 May 2026; the IMCO and LIBE committees approved that agreement on 2 June 2026; and the Parliament’s plenary vote is scheduled for 16 June 2026. After the plenary vote, the Council must still formally endorse the text, which then undergoes legal-linguistic revision before publication in the Official Journal and entry into force three days later.

The headline changes are the staggered high-risk deferrals above, the targeted Article 50(2) grace period, and a new prohibition added to Article 5 covering AI-generated non-consensual intimate imagery (“nudifiers”) and child sexual abuse material. The key practical point for this timeline: none of the deferred dates bind until the Omnibus is published in the Official Journal. Until that happens, the original 2 August 2026 high-risk date remains the law on the books, and organisations are advised to keep preparing against it.

Penalties for missing a deadline

Article 99 sets three tiers of administrative fines, live since 2 August 2025. For undertakings, the figure applied is the higher of the fixed amount or the percentage of total worldwide annual turnover.

SME and start-up protection (Art. 99(6)). For SMEs and start-ups the rule inverts: each fine is capped at the lower of the fixed amount or the percentage — not the higher. For a company with €2 million in annual turnover, a tier-two Article 50 violation is capped at 3% of turnover (€60,000), not €15 million. If you are a small operator, this is the single most important sentence on the page.

Frequently asked questions

What is the next AI Act deadline?

2 August 2026 — the date the Article 50 transparency obligations become applicable. It is the next deadline that binds new substantive obligations, and it has not been deferred by the Digital Omnibus.

Has the AI Act high-risk deadline been postponed?

It is being postponed, but it is not yet final. The Digital Omnibus would move stand-alone Annex III high-risk obligations from 2 August 2026 to 2 December 2027, and Annex I product-embedded systems from 2 August 2027 to 2 August 2028. These dates bind only once the Omnibus is published in the Official Journal — expected in 2026 but not confirmed as of 16 June 2026.

Does Article 50 apply to my chatbot?

If you provide a system intended to interact directly with people, Article 50(1) requires you to ensure users are informed they are dealing with an AI — unless that is obvious from the circumstances and context of use. Most customer-facing chatbots fall within scope; our guide to chatbot and AI-interaction disclosure under Article 50(1) covers exactly how to comply.

When do AI Act fines start applying?

The penalty framework in Chapter XII has been applicable since 2 August 2025. Fines attach to each obligation as that obligation becomes applicable on its own date in the timeline.

Is the AI literacy requirement already in force?

Yes. The Article 4 AI literacy duty has applied since 2 February 2025, alongside the Article 5 prohibitions.

Key takeaways

• The AI Act applies in phases under Article 113 — there is no single “go-live” date.
• The next binding deadline is Article 50 transparency on 2 August 2026, and it has not been deferred.
• The Digital Omnibus defers high-risk obligations to 2 December 2027 (Annex III) and 2 August 2028 (Annex I), but only once it is published in the Official Journal.
• A narrow grace period gives legacy generative AI until 2 December 2026 for Article 50(2) marking only.
• Article 99 fines run to €35M/7%, €15M/3% and €7.5M/1% — but for SMEs each cap is the lower figure, not the higher.

Bookmark this AI Act timeline — we revise it the moment the Digital Omnibus is published in the Official Journal.