Over the past few months a lot of confusion has spread among SMEs about what was actually “postponed” in the AI Act, following the political debate around the so-called AI Omnibus. The short version: Article 50’s transparency obligations have not been postponed. They remain anchored to 2 August 2026. What is shifting concerns other parts of the regulation, not Article 50 as a whole.
If your business uses any AI tools — even just a customer-support chatbot or a content-generation system — this article clarifies what expires when, and what to do to be ready.
What Article 50 actually says
AI Act Article 50 introduces transparency obligations: people must be able to know when they are interacting with an AI system, when content (text, image, audio, video) is artificially generated or manipulated, and in some cases synthetic content must be marked in a machine-readable way.
The key dates to keep in mind are:
- 2 August 2026 — remains the reference date for Article 50 transparency obligations. This is the deadline that affects most SMEs using generative AI, chatbots, or tools that produce content for customers or employees.
- 2 December 2026 — for Article 50(2), relating to machine-readable marking/watermarking of synthetic content, an additional period applies to systems that already existed as of this date.
- 2 December 2027 — is the date relevant to obligations for high-risk Annex III systems (not Article 50).
- 2 August 2028 — is the date relevant to AI embedded in products regulated under Annex I (again, not Article 50).
The key point is that these last three dates do not concern the general Article 50 transparency obligation that applies to most SMEs. They concern more specific categories (high-risk, regulated products, technical marking for pre-existing systems).
Where the “AI Omnibus” confusion comes from
The political discussion around the “Omnibus” package proposed — and in some cases reached political agreement on — longer timelines for specific categories of obligations, particularly for high-risk Annex III systems and for AI embedded in Annex I products. This generated simplified headlines along the lines of “the AI Act has been postponed,” which do not accurately reflect the situation.
For an SME using AI in a “normal” way (customer support, marketing, internal productivity), the practical takeaway is: don’t assume everything has been postponed. Check which obligation actually applies to your specific case before relaxing about deadlines.
What SMEs should do now
Regardless of how the political debate on high-risk timelines evolves, there is readiness work that makes sense to start now, because the 2 August 2026 transparency deadline is approaching:
- Identify where you use AI. First you need an inventory: which generative AI tools, chatbots, or automated systems your business uses, whether customer-facing or internal.
- Prepare Article 50 transparency notices. For every customer-facing use that generates or manipulates content, assess whether a notice is needed to communicate the artificial nature of the content or interaction.
- Document internal responsibilities. Who is responsible for keeping the AI inventory up to date? Who decides when a new transparency notice is needed?
- Keep evidence. Screenshots, published notice text, update dates: a minimal evidence log makes it much easier to demonstrate due diligence if questioned.
This kind of preparatory work — inventory, notices, responsibilities, evidence — is exactly the documentation and readiness support we offer through the Free Assessment and the Compliance Pack.
What we are not
For clarity: EU AI Passport provides readiness and documentation support, not legal advice or an official conformity assessment. Every business situation is different, and for complex or high-risk cases we always recommend a review with a qualified legal advisor.
This article is for general information and is not legal advice.