Fully on top of the EU AI Act
Current insights, clear guidance and practical tools that help you respond quickly and correctly to what changes.
Trending tools
New developments?
The developments drawing the most attention and searches right now.
Highest point in a year. Interest peaks each time toward the August deadlines.
Digital Omnibus: Council agrees, deadlines pushed back
The Council gave the green light for a lighter AI Act. Stand-alone high-risk systems get until 2 December 2027, systems in products until 2 August 2028. The text applies after publication in the Official Journal.
Source: Raad van de EU
The labelling duty (Art. 50) goes ahead
Transparency for AI images, video, audio and deepfakes starts on 2 August 2026. Systems already on the market get until 2 December 2026 for the machine-readable marking.
Source: artificialintelligenceact.eu
A new ban added
The Omnibus bans AI that creates non-consensual intimate imagery or child sexual abuse material.
Source: Raad van de EU
GPAI code of practice finalised
The code of practice for general-purpose AI models, like the ones behind ChatGPT, is ready. The GPAI obligations have applied since August 2025.
Source: Europese Commissie
Most-searched questions
The questions people type most this week. Each with a short, checked answer and a source. Since the Omnibus, delay and labelling top the list.
1Is the AI Act delayed? Current›
Partly. The Digital Omnibus (Council agreement on 29 June 2026) mainly delays the high-risk obligations: stand-alone systems to 2 December 2027, systems in products to 2 August 2028. The rest stays. Prohibited practices apply since 2025, GPAI since August 2025, and the labelling duty from 2 August 2026. So "the whole AI Act is delayed" is not correct.
Source: Raad van de EU · Implementatietijdlijn
2Do I need to label my AI content? Popular›
It depends on your role. The provider of the AI tool must mark the output machine-readably. If you yourself publish realistic AI images, video, audio or deepfakes, you disclose that visibly and clearly. It applies from 2 August 2026. Systems already on the market get until 2 December 2026. Not every AI output is covered: it concerns synthetic or edited media and visible AI conversations.
Do it now: free AI Label Tool · Source: Artikel 50
3When does the EU AI Act apply? Popular›
In phases. Prohibited practices and AI literacy since 2 February 2025. GPAI and governance since 2 August 2025. The labelling duty (Article 50) from 2 August 2026. The high-risk obligations move, via the Omnibus, to 2 December 2027 and 2 August 2028.
Source: Implementatietijdlijn
4What is the EU AI Act? Popular›
The AI Act (Regulation (EU) 2024/1689) is the world's first broad AI law. It applies across the EU and classifies AI by risk. The higher the risk, the stricter the rules.
Source: High-level summary · EUR-Lex
5Who does the law apply to? ›
To providers who build or sell AI, and to deployers who use AI professionally. Also to companies outside the EU, as soon as their AI output lands in the EU. Small organisations have duties too, such as AI literacy and disclosing the AI content they publish.
Source: Verordening (EU) 2024/1689
6What risk categories are there? ›
Four. Unacceptable risk is banned, like social scoring. High risk is allowed under strict requirements, like AI for recruitment or critical infrastructure. Limited risk needs transparency, like chatbots and AI images. Minimal risk has no extra duties. Most everyday AI sits in those last two.
Source: High-level summary
7How high are the fines? ›
Up to € 35 million or 7% of global annual turnover for prohibited practices. For other breaches the caps are lower. Supervision sits with the European AI Office (for GPAI) and the national authorities.
Source: Europees Parlement
8Who supervises in Belgium? BE›
The BIPT is the main market surveillance authority for the AI Act in Belgium. Alongside it, the Data Protection Authority (privacy), the FAMHP (medicines) and the FSMA (financial sector), among others, are involved. For fundamental rights, Belgium designated a record number of bodies.
Source: National implementation plans
Factcheck
Since the Omnibus coverage in particular, half-truths are going around. On the left the claim you often hear, on the right what it actually says. With a source.
"The whole AI Act is delayed."
No. The Omnibus mainly delays high-risk, to December 2027 and August 2028. Prohibited practices, GPAI and the labelling duty stay in place.
"The AI Act bans AI."
The law works by risk. Only a short list of 'unacceptable' uses is banned. Most AI falls under minimal or limited risk and is simply allowed.
"Only big tech needs to worry."
An SME that uses AI has duties too: AI literacy, and disclosing the AI content it publishes. You don't have to build the AI yourself.
"You must label every AI output."
Only certain cases: synthetic or edited images, video and audio, deepfakes, and visible AI conversations. Internal use or non-realistic output usually not.
"The AI Act stops at the EU border."
The law works extraterritorially. As soon as your AI output is used in the EU, you can fall under it. Even from abroad.
The essentials
The constant core, apart from the week's news. The AI Act classifies AI by risk and ties duties to it.
Timeline (updated after the Digital Omnibus)
Everything on the EU AI Act
Hand-picked sources, tools and analyses. Each with a label: official, Belgian or commentary. Search at the top or browse by category.
No results for that search.
About this site
eu-ai-act.be is an independent, free hub on the EU AI Act. We are not part of the EU or any government. We follow the official sources and the sharpest analyses, and update the state of play and the most-searched questions every week. General information, not legal advice.
References
- Council of the EU, Council gives final green light to simplify and streamline AI rules (29 June 2026).
- Implementation timeline, Article 50 en High-level summary, artificialintelligenceact.eu.
- Europese Commissie, AI Act · Shaping Europe's digital future en GPAI Code of Practice.
- Verordening (EU) 2024/1689, EUR-Lex.
- European Parliament, EU AI Act: first regulation on artificial intelligence.
Dates and figures were updated after the Digital Omnibus (June 2026) and are reviewed every week. The published regulation is binding.