The transparency obligations started to apply on 2 August 2026, but the deadlines for high-risk systems were postponed by the Digital Omnibus. The real position on 20 August 2026, for companies that use AI.
2 August 2026 was long presented as the “big day” of the AI Act: the moment when the rules for high-risk systems and the transparency obligations would, in the main, start to apply. If you read about it in 2025 or at the start of 2026, you probably have a picture that no longer matches reality. Below is the position on 20 August 2026, with official sources.
This is not legal advice. If a decision in your company depends on these deadlines, check them with a lawyer.
In short
- The transparency obligations in Article 50 apply from 2 August 2026, with a grace period until 2 December 2026 for the machine-readable marking of generated content, for systems already on the market.
- High-risk systems in Annex III (for example uses in recruitment, education, credit scoring) were postponed from 2 August 2026 to 2 December 2027.
- High-risk systems embedded in regulated products (Annex I) apply from 2 August 2028.
- The prohibitions and AI literacy have applied since 2 February 2025; the rules for general-purpose models since 2 August 2025.
What moved the deadlines is the so-called “Digital Omnibus” on AI.
What the Digital Omnibus on AI is
The European Commission proposed a simplification package at the end of 2025, and the AI part was politically agreed on 7 May 2026, according to the Commission’s page on the AI regulatory framework. The text entered into force on 27 July 2026. It is published as Regulation (EU) 2026/1744 and amends the AI Act: hence the postponement of the deadlines for high-risk systems.
So it is no longer merely a proposal: on the date of this article, the postponement is in force.
What already applies
To keep track, here is the timeline in stages, as the Commission presents it:
- 1 August 2024: the AI Act entered into force.
- 2 February 2025: the prohibited practices and the AI literacy obligation apply.
- 2 August 2025: the governance rules and the obligations for general-purpose models apply.
- 2 August 2026: the Act becomes applicable, “with some exceptions”; the transparency rules come into effect, and the AI Office and Member State authorities take over implementation and enforcement.
- 2 December 2026: a new prohibited practice added by the Omnibus (generating non-consensual sexual content and child sexual abuse material) and the grace period for marking generated content in systems already on the market.
- 2 December 2027: high-risk systems in Annex III.
- 2 August 2028: high-risk systems in Annex I.
Transparency: what it means for a small company
Article 50 distinguishes between whoever provides an AI system and whoever uses it (in the language of the law, the “deployer”). For a company using ready-made tools, what matters most is what the law asks of the provider and what stays on your shoulders:
- Interaction with people. The provider of a system intended to interact directly with people (for example a chatbot) must design it so that people are informed they are talking to an AI system, unless that is obvious from the context. Even if the formal obligation is the provider’s, it is good practice to say it clearly yourself in the website chat, that an automated assistant is replying.
- Artificially generated or altered content. Providers of generative systems must mark outputs in a machine-readable format. This is where the grace period until 2 December 2026 applies, for systems placed on the market before 2 August 2026.
- Deepfakes and texts published to inform the public. Whoever uses an AI system to generate or manipulate images, audio or video that look authentic (deepfakes) must disclose that they are artificially generated. Similar obligations exist for AI-generated text published to inform the public on matters of public interest, with exceptions, for example where there is human editorial control.
- Emotion recognition and biometric categorisation. Whoever uses them must inform the people exposed.
What exactly this would mean for the images in your ads, your blog posts or your marketing materials depends on the case and on the exceptions. This is where a legal opinion is worth having, rather than a general rule.
AI literacy: relaxed somewhat
The obligation in Article 4 has applied since 2 February 2025. The Omnibus reworded it: providers and those who use AI must take measures to support the development of AI literacy among their staff, and the text specifies that it does not require guaranteeing a specific level of literacy for every individual. The Commission is to publish practical examples. As a small company you still have something to do: train the people who work with AI and note what you did. Check the final wording with a lawyer.
High risk: postponed, not forgotten
The 2 December 2027 deadline for Annex III does not mean you can forget the subject. A few reasons:
- If you use AI for decisions about people, for example in CV screening, you need to know in good time whether you fall within the high-risk area.
- The GDPR already applies: automated processing with significant effects on a person has limits (Article 22) and requires proper information.
- The high-risk requirements mainly concern the provider, but whoever uses the system also has obligations (human oversight, keeping logs, informing people).
What to do now, in small steps
- Make a list of the AI tools used in the company: chatbots, image generators, assistants, AI features in existing software.
- For each one, note what it is used for and what data it receives.
- Mark what talks to customers and what publishes something on the company’s behalf: transparency applies there.
- Mark what influences decisions about people (recruitment, credit, access to services): watch Annex III there.
- Train the team and keep a short note of it.
- Come back to the list when there is news. The deadlines have moved once; new clarifications (guidelines, codes of practice) may appear.
At ME Digital we help small companies inventory and configure their AI tools; see AI implementation or get in touch. Interpreting the law stays with your own lawyer.