Guides

EU AI Act from 2 August 2026: what now applies, what was delayed, and what to check

Most of the EU AI Act became applicable on 2 August 2026, including the Commission's power to fine general-purpose AI providers and new transparency duties. High-risk rules were pushed back to 2027 and 2028.

Abstract illustration of a dotted route through waypoints over a contour map, in amber and gold.
Abstract illustration of a dotted route through waypoints over a contour map, in amber and gold.

The EU AI Act became generally applicable on 2 August 2026. From that date the European Commission can enforce the obligations on providers of general-purpose AI (GPAI) models, including with fines, and the Act’s transparency rules for chatbots and AI-generated content apply. The main obligations for high-risk AI systems in sensitive areas such as employment and education, originally due on the same day, do not yet apply: the “AI Omnibus” amending regulation moved them to 2 December 2027, and set 2 August 2028 for high-risk AI built into regulated products.

This guide is general information, not legal advice. If the AI Act affects your organisation, check the official texts and take qualified legal advice.

What changed on 2 August

According to the Commission, from 2 August 2026 the AI Office and national authorities are responsible for implementing, supervising and enforcing the Act. Two areas matter most for developers and businesses.

General-purpose AI models. GPAI obligations have applied since 2 August 2025, but Article 101, which lets the Commission fine providers, applies only from this year. Fines can reach 3% of annual worldwide turnover or €15 million, whichever is higher, where a provider intentionally or negligently breaches the rules, fails to answer a request for information, fails to comply with a requested measure, or does not give the Commission access to a model for evaluation. Providers of models placed on the market before 2 August 2025 have until 2 August 2027 to comply (Article 111(3)).

Transparency (Article 50). Four obligations now apply:

  • Providers must make sure people are told when they are interacting with an AI system, unless that is obvious.
  • Providers of systems that generate synthetic audio, images, video or text must mark outputs in a machine-readable format so they can be detected as AI-generated.
  • Deployers of emotion recognition or biometric categorisation systems must inform the people exposed to them.
  • Deployers must disclose deepfakes, and AI-generated text published to inform the public on matters of public interest, unless that text has undergone human review or editorial control and a person or organisation holds editorial responsibility for it.

Information must be given clearly, at the latest at the first interaction or exposure. Breaches of Article 50 can attract fines of up to €15 million or 3% of worldwide turnover, whichever is higher; for SMEs and start-ups the lower of the two applies (Article 99).

The timeline after the AI Omnibus

The amending regulation, Regulation (EU) 2026/1744, was proposed on 19 November 2025, agreed politically on 7 May 2026, voted by Parliament on 16 June, adopted by the Council on 29 June and published in the Official Journal on 24 July. It entered into force on 27 July 2026. The consolidated dates now read:

Obligation Applies from
Prohibitions 1–8 and AI literacy 2 February 2025
GPAI model obligations 2 August 2025
General application; Commission fines for GPAI providers; Article 50 transparency 2 August 2026
Article 50(2) marking for generative systems already on the market before 2 August 2026 2 December 2026
New prohibition on AI that generates non-consensual intimate imagery or child sexual abuse material 2 December 2026
At least one national AI regulatory sandbox per member state 2 August 2027
GPAI models placed on the market before 2 August 2025 2 August 2027
High-risk systems in Annex III areas (for example employment, education, biometrics) 2 December 2027
High-risk systems in regulated products under Annex I 2 August 2028

The Omnibus also softened the AI literacy duty in Article 4: providers and deployers must take measures to support their staff’s AI literacy, but are no longer required to guarantee any specific level for each individual.

Step by step: what to check now

  1. List your AI systems and your role in each. Article 50 places some duties on providers (who build or place a system on the market) and others on deployers (who use it). Many organisations are both.
  2. Chatbots and assistants. If people interact directly with your system, check that they are told it is AI at or before the first interaction.
  3. Generative AI products. If you provide a system that generates content, check how its outputs are marked and made detectable. If it was on the EU market before 2 August 2026, the deadline for this specific duty is 2 December 2026 (Article 111(4)). The Commission says outputs generated before 2 August 2026 do not need to be marked retroactively.
  4. Publishing and media workflows. If you deploy AI to produce deepfakes or public-interest text, check your labelling, and whether your editorial review meets the human-review exception.
  5. Read the official guidance. The Commission published guidelines on Article 50 on 20 July 2026. A voluntary Code of Practice on Transparency of AI-generated Content, published on 10 June, was judged adequate by the Commission on 8 July. Signing is done by emailing a completed form to the AI Office; the Commission says not signing is not in itself non-compliance, but non-signatories must show compliance by other means.
  6. GPAI model providers. Review your technical documentation and the Commission’s guidelines for GPAI providers, since requests for information and evaluations now carry the possibility of fines.
  7. Keep high-risk work going. The delay moves the deadlines; it does not remove the requirements.

What it means

For most companies the practical change this month is transparency rather than high-risk compliance. The Omnibus gave Annex III high-risk systems roughly 16 more months, but chatbot disclosure, content marking and deepfake labelling are live now, with a short grace period only for marking by existing generative systems. For model providers, the AI Office now has enforcement powers to go with the obligations that took effect a year ago, which makes documentation and responsiveness to Commission requests a compliance matter rather than a courtesy.

Sources

See All

Type to search. ↑ ↓ to navigate, ↵ to open.