The AI Act arrives in stages. Some duties started in 2025. The date that matters for most consumer facing businesses is 2 August 2026, because that is when the transparency obligations in Article 50 apply. If you run a chatbot or generate content for EU users, this is your date.
What switches on for ordinary businesses
- Tell people they are talking to AI. The chatbot interaction notice under Article 50(1).
- Label deepfakes. Disclose artificially generated or manipulated image, audio, and video under Article 50(4).
- Flag AI generated content. Mark synthetic media, and disclose emotion recognition or biometric categorisation.
What the simplification package did not do
There was a lot of noise about a Commission simplification package easing the Act. It is important to be precise: the transparency obligations in Article 50 were not postponed. They apply on 2 August 2026 as planned. The only related easing is narrow: the machine readable marking of synthetic output under Article 50(2) gives systems already on the market before that date until 2 December 2026 to comply. The visible duties, telling users about a chatbot and labelling deepfakes, are live on 2 August.
A simple way to be ready
You do not need a compliance project to show a disclosure notice. The gap for most sites is small: a clear, timely, localized notice on the AI touchpoint, plus a record that it ran. A hosted widget handles all of that from one snippet and updates itself if the recommended wording changes after August.
Common questions
Was Article 50 delayed to 2027?
No. Article 50 transparency applies from 2 August 2026. The 2027 dates in the timeline relate to other parts of the Act, not the transparency notices.
What is the 2 December 2026 date?
It is the grace window for adding machine readable marking to synthetic output, but only for systems already on the market before 2 August 2026.
Be ready before 2 August 2026
Drop in the widget now and the Article 50 notice is live, localized, and self updating well before the deadline.
Start free