Key takeaways
- High-risk deadlines moved to 2 December 2027 (Annex III) and 2 August 2028 (Annex I).
- The delay doesn't cover Article 50 transparency, which applies from 2 August 2026, or the rules for general-purpose AI models.
- Systems already on the market have until 2 December 2026 to mark AI-generated content. New systems must comply from launch.
- A simple AI inventory and a clear disclosure on chatbots are cheap steps that hold up whatever the timeline.
What changed
The Council formally adopted the AI Omnibus on 29 June 2026, according to iubenda’s analysis, and Hunton reports that Regulation (EU) 2026/1744 entered into force on 27 July 2026. The headline effect is more time for high-risk AI systems. Hunton also notes that the Omnibus adds a ban on AI systems designed to generate non-consensual intimate imagery and child sexual abuse material.
This is general information, not legal advice. Check your own position with a qualified adviser.
The new timeline at a glance
| Obligation | Date |
|---|---|
| Article 50 transparency duties | 2 August 2026 (unchanged) |
| Content-marking (watermarking) for systems already on the market before 2 August 2026 | 2 December 2026 |
| Content-marking for systems placed on the market after 2 August 2026 | From launch, no grace period |
| High-risk systems listed in Annex III (for example employment and education) | 2 December 2027 |
| High-risk AI embedded in regulated products (Annex I) | 2 August 2028 |
| Obligations for general-purpose AI models (Articles 51 to 55) | Original schedule, unaffected |
Sources vary in how much detail they give, so check the dates against the official text before you build a compliance plan on them.
What still applies today
The delay is for high-risk systems. A lot of small and mid-sized businesses use AI in lower-risk ways, and those fall under the transparency rules instead. In short, Article 50 is about people knowing when they’re dealing with AI. If your website chatbot, voice assistant or content tool reaches people in the EU, plan for:
- Clear disclosure that someone is talking to an AI system, unless it’s obvious.
- Marking of AI-generated or AI-manipulated content where the rules require it.
- Human oversight for decisions that affect people. That’s good practice with or without the Act.
The Omnibus didn’t touch other laws such as the GDPR, so they still apply to any personal data your AI systems handle.
Who is affected outside the EU
The AI Act can reach providers and deployers outside the EU when their AI systems are used in the EU. So if you’re a UK, US or Gulf business serving EU customers through a chatbot or content tool, check your exposure even though the law comes from Brussels. If you work across regions, our UK and Europe consultancy page explains how we approach it.
A practical checklist
- List your AI systems. Include chatbots, scoring tools, content generators and anything a vendor supplies.
- Classify each one as likely high-risk or not, and note which ones talk to the public.
- Add disclosure to chatbots and voice assistants.
- Label AI-generated media in your content workflow, and keep a human approval step.
- Ask your vendors how they’ll meet the new dates and what documentation they’ll give you.
- Set a review date for December 2026.
These steps make later compliance cheaper too, because most of the work is simply knowing what AI you use and where.
How TendWise AI approaches this
We build agents and chatbots with disclosure, permissions, logging and human handoff from the start, which makes them easier to adjust when rules change. If you’d like a second opinion on a setup you already have, begin with an AI strategy review, or read how we build AI chatbots that hand over to people when they need to.
Frequently asked questions
When does the EU AI Act Digital Omnibus take effect?
Did the Omnibus delay chatbot transparency rules?
Does the EU AI Act apply to non-EU companies?
What is the easiest first step for a small business?
Sources
- iubenda: AI Omnibus adopted, what still applies from August 2
- Hunton: EU Digital Omnibus on AI enters into force
- Sidley: EU lawmakers reach provisional agreement to delay key AI Act obligations