Marking deadline for existing systems in —
The AI Act is in force. Article 50 already requires every company whose AI talks to people, or generates text, images, audio or video, to tell them clearly. Providers of AI models and high-risk systems established outside the Union need an authorised representative inside it. We give you the notices, the register and the representative, in eight languages, at a fixed price.
Europe Services, SE — Prague, Czech Republic. Active since 2014. Article 27, Article 16, CRA and AI Act mandates signed for companies in 24 countries.
A first orientation, not legal advice. The classification of your system stays your decision.
The AI Act follows the AI, not the server. If people in the European Union meet your AI system, or use its output, the rules reach you, wherever the company is. What you owe depends on your role: the provider who builds or sells the system, or the deployer who uses it.
| Situation | What the law asks | What we do |
|---|---|---|
| Website or app with an AI chatbot or voice assistant | Article 50(1): tell people they are talking to an AI, unless it is obvious. | We write the notice in eight languages and keep every version. |
| AI-generated images, video, audio or text published to the public | Article 50(2) and (4): mark it as artificially generated; disclose deep fakes. | Notices, marking guidance and a public badge. |
| Provider of a general-purpose AI model, outside the EU | Article 54: appoint an authorised representative in the Union by written mandate. | We are the representative and hold your documentation. |
| Provider of a high-risk AI system, outside the EU | Article 22: authorised representative, from 2 December 2027 for Annex III systems. | We sign the mandate now, so the address is ready. |
| Company that only uses internal AI tools, no contact with the public | Few duties under Article 50; AI literacy and good governance still expected. | We tell you honestly that you may not need us. |
Almost every question about the AI Act comes down to two facts about your system. Get them right and the rest follows.
A chatbot on your site, a voice on your phone line, an AI that writes to customers or creates images they will see. If people meet it or its output, Article 50 applies to you as deployer or provider.
If you build or sell the AI under your name, you are its provider. If it decides on jobs, credit, education, access to services or safety, it may be high-risk under Annex III, with the heaviest duties.
We record your answers in the mandate. The classification is yours; we make sure the paperwork matches it.
A lot of what is being sold as "AI Act compliance" is not due yet. Regulation (EU) 2026/1744, the Digital Omnibus on AI, moved several dates. We would rather tell you which ones.
Prohibited practices since February 2025. General-purpose AI models since 2 August 2025. Article 50 transparency since 2 August 2026, with marking of content from systems already on the market by 2 December 2026.
High-risk systems listed in Annex III (employment, credit, education, biometrics, critical infrastructure): 2 December 2027. High-risk AI inside regulated products (Annex I): 2 August 2028.
The practical catch: an authority does not ask whether you knew the date. It asks where your notice is, since when, and who answers for it. That is what we keep.
Not a PDF of generic advice. A set of texts in your name, a record that proves them, and a person who answers when someone asks.
The chatbot notice, the AI-content label, the deep-fake disclosure and, where needed, the emotion-recognition notice. Written for your systems, in English plus up to seven languages.
Where each notice goes: first message of the chat, under the image, in the video, in the email footer. The law cares about the moment people meet the AI.
Each change of each notice, with its date. If an authority asks what users saw in March, you have the answer in one minute.
What each system does, whether you are provider or deployer, where people meet it, which notice applies, when it was last reviewed. Kept by us, available to you and to an authority.
For providers outside the Union, the authorised representative is the address the AI Office and national authorities write to. We sign the mandate, hold the documentation for ten years and answer.
A public verification page under your AI27- code, and a badge for your site. Buyers and partners can check it themselves.
Wider than people expect. A machine-based system that works with some autonomy and infers from its inputs how to generate outputs such as predictions, content, recommendations or decisions. A chatbot built on someone else's model is an AI system; so is a scoring model in your back office.
Outside: systems used only for military, defence or national security, pure research before placing on the market, and personal non-professional use. Free and open-source models have lighter duties, but not when they are placed on the market as high-risk or used in prohibited practices.
| Date | Who | What applies | Enforced by |
|---|---|---|---|
| 24h | 2 Feb 2025 · everyone | Prohibited practices (Article 5) and AI literacy (Article 4). | National authorities |
| 72h | 2 Aug 2025 · model providers | General-purpose AI models: documentation, copyright policy, training-data summary, authorised representative for non-EU providers. | AI Office |
| 14d | 2 Aug 2026 · deployers and providers | Article 50 transparency: chatbots, AI-generated content, deep fakes, emotion recognition. Content from systems already on the market: by 2 Dec 2026. | Market surveillance authorities |
| 1m | 2 Dec 2027 · high-risk | Annex III high-risk systems: risk management, data governance, logging, human oversight, registration, authorised representative. Annex I systems: 2 Aug 2028. | Market surveillance authorities |
Dates as amended by Regulation (EU) 2026/1744 (Digital Omnibus on AI), in force since 27 July 2026.
Companies often think a GDPR representative covers AI. It does not. The three rules ask different things of different people.
| AI Act | GDPR | CRA | |
|---|---|---|---|
| Who is bound | Provider or deployer of the AI | Controller of personal data | Manufacturer of connected products |
| What it asks first | Transparency to people; representative for non-EU providers | Representative for non-EU controllers (Article 27) | 24-hour vulnerability reporting |
| First deadline | 24h | 24h | 72h |
| Enforced by | AI Office and market surveillance authorities | Data protection authorities | CSIRTs and ENISA |
| Applies from | 2025 – 2028, in stages | In force | 11 September 2026 |
One chatbot can trigger all three: it talks to people (AI Act), it processes their data (GDPR), and it runs in an app with a data connection (CRA). We handle each as a separate mandate on the same account.
Transparency breaches sit in the band of up to €15 million or 3% of worldwide annual turnover. Prohibited practices reach €35 million or 7%. For small and medium companies the lower of the two amounts applies, which is still more than any plan on this page.
The commercial damage comes first. Platforms, app stores and enterprise buyers already ask for AI disclosures in their questionnaires. A missing label is a lost contract before it is a fine.
Three moments of the same duty: the code that runs the AI, the notice people read, and the person who answers for it.
One price, one renewal, no hourly billing. Large firms quote €2,000 – €5,000 a year for the representative alone.
AI TRANSPARENCY
€290
Renewal €240
Article 50 · deployers and providers
REGISTER + TRANSPARENCY
€590
Renewal €490
Companies with several AI systems
AI ACT REPRESENTATIVE
€1490
Renewal €1190
Providers outside the Union · per model or system
We do not sell audits, conformity assessments, AI testing or model evaluations, and we are not a notified body. When you need one, we tell you.
We have followed the text since the proposal and revise this page each time the dates move, as they did in July 2026.
| Date | Milestone | What changed |
|---|---|---|
| 2024-10-23 | Adoption | Regulation (EU) 2024/1689 published on 12 July 2024. The world's first horizontal AI law. |
| 2024-12-10 | Entry into force | 1 August 2024. Nothing enforceable yet; the clock for every later stage starts here. |
| 2026-06-11 | Prohibitions | 2 February 2025: prohibited practices and AI literacy apply. |
| 2026-09-11 ← | GPAI and Article 50 | 2 August 2025 for general-purpose models; 2 August 2026 for transparency to people. |
| 2027-12-11 | Digital Omnibus | Regulation (EU) 2026/1744, in force 27 July 2026: high-risk deadlines moved to December 2027 and August 2028; Article 50 unchanged. |
Most arguments about the AI Act are arguments about five or six definitions. These are the ones that decide what you owe.
Everything on this page comes from the regulation and the official pages of the Commission. Where we interpret, we say so.
The consolidated text of the AI Act on EUR-Lex, in all official languages.
The Commission's page on the regulatory framework and its timeline.
What the Article 54 mandate covers for general-purpose AI models.
The full terms of the three plans, in English and Italian.
Two representatives, two different duties.
Check any certificate we issued, with its code.
Page maintained continuously since the regulation entered into force. Last revision: 26 September 2026.
Tell us which AI you use and where people meet it. We answer with the one thing you need first: whether the AI Act asks you for anything at all.
Annual fee, paid in advance. You can cancel before each renewal.
From the second year the renewal is lower: €240, €490 and €1190 a year.
Where is the company established?
No VAT for customers outside the EU. Inside the EU with a valid VAT number: reverse charge.
Tell us which AI you use or sell, and where. We answer with whether the AI Act reaches you at all.
We reply within one business day, in your language. No sales sequence, no newsletter.
One working day for an answer. No payment at this step.
Payment received. One ten-minute form is missing: your AI systems and who answers for them.
REP27