Skip to main content
Praxikon
Agent check

Free template · Word · English and Dutch

AI transparency notice template for Article 50 of the EU AI Act

Article 50 of the EU AI Act sets transparency obligations for certain AI systems. Providers design AI systems that interact directly with people so that those people are informed they are dealing with AI, and mark generated content in a machine-readable format. Deployers inform people exposed to emotion recognition or biometric categorisation and disclose deep fakes and certain AI-generated text. These obligations have applied since 2 August 2026; the information must be given at the latest at the first interaction or exposure.

Last checked against the law
Editor
, jurist, privacy and AI
Version and template ID
2.0 · praxikon:template:ai-transparency-notice
Legal basis
Article 50 and Article 111(4) of Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744

Who is this template for?

  • Owner of a chatbot, voicebot or AI agent in customer contactNotice when people interact directly with AI (Article 50(1), provider) and timing and form (Article 50(5))
  • Communications, marketing and editorial teamsLabels for deep fakes and for published AI-generated text on matters of public interest (Article 50(4), deployer)
  • Product and engineering teams that provide an AI system generating contentMachine-readable marking (Article 50(2)) and the 2 December 2026 deadline for existing systems (Article 111(4))
  • Data protection officer or privacy officerEmotion recognition and biometric categorisation (Article 50(3)) alongside Articles 9, 13, 22 and 35 GDPR
  • Procurement and legalDetermining the role of provider or deployer (Article 3(3) and (4)) and agreements with the supplier on the notice
  • Compliance officer or AI governance ownerEvidence log per application and channel; maximum fine under Article 99(4)(g)

What is inside

  • Article 50 in one table: for each paragraph who carries the obligation, when it applies and since when (start of the kit)
  • Decision tree for paragraphs 1 to 4, with six examples to determine your role as provider or deployer (chapter 1)
  • Timing and form under paragraph 5, the recommendations of the Commission guidelines and a go-live checklist (chapter 2)
  • Notice and label texts in English, with a Dutch example version, for chat, voice, email by an AI agent, images, deep fakes, published text and age estimation (chapter 3)
  • Provider section on machine-readable marking, with what the code of practice asks per type of output and a checklist (chapter 4)
  • What a notice cannot fix: the prohibitions of Article 5, including the new prohibitions from 2 December 2026 (chapter 5)
  • Link with the GDPR per situation, with a ready-to-use paragraph for your privacy notice (chapter 6)
  • Evidence log with fictional examples and an overview of what to keep per outcome (chapter 7)

How to use the template

  1. List every application that talks to people or creates content: chat, telephone, email, images, video, audio and text, including your suppliers' applications.
  2. Determine your role and the relevant paragraph for each application with the decision tree in chapter 1, or online with the Article 50 decision tree.
  3. Pick the matching text from chapter 3, fill in the words in [brackets], translate it into the languages of your audience and place the notice at the moment and in the form set out in chapter 2.
  4. If you provide a system that generates content, work through the machine-readable marking checklist in chapter 4.
  5. Record in the evidence log of chapter 7 what is live for each application and channel, since when and who monitors it.
  6. Check every quarter, and after every change of supplier, house style or call script, that the notice is still in place.

Common mistakes

  • Putting the notice only in the terms and conditions or the privacy notice. Paragraph 5 requires the information to be clear and distinguishable, at the latest at the first interaction or exposure.
  • Treating Article 50 as a future obligation. It has applied since 2 August 2026; only the machine-readable marking of paragraph 2 has a deadline of 2 December 2026, for systems placed on the market before 2 August 2026.
  • Assuming the supplier takes care of everything. If you build a chatbot on someone else's model and deploy it under your own name, you may be the provider yourself; the labelling of deep fakes and AI-generated text (paragraph 4) always rests with the deployer.
  • Labelling every AI image visibly, or nothing at all. Under Article 50 a label is required for deep fakes and for AI-generated text published to inform the public on matters of public interest, unless that text has undergone human review or editorial control and someone holds editorial responsibility; other law, such as the rules against misleading advertising, may also require a disclosure.
  • Trying to fix emotion recognition in the workplace or in education with a notice. It has been prohibited since 2 February 2025, except for medical or safety reasons (Article 5(1)(f)).

Prefer to work online? AI disclosure texts to copy, per channel · Article 50 decision tree

When do you need legal advice?

  • Your role is not clear-cut. You build a chatbot or content generator on a supplier’s model and deploy it under your own name. You may then be the provider for paragraph 1 or 2, and not only the deployer.
  • You want to rely on an exception. For example that it is obvious to the user that they are talking to AI (paragraph 1, provider’s obligation), that an AI-generated text has undergone editorial control (paragraph 4, deployer’s obligation), or that a limited disclosure suffices for an evidently artistic or satirical work. These exceptions are narrow and you must be able to substantiate them.
  • The system recognises emotions or categorises people on biometric features. Paragraph 3 then applies to you as deployer, alongside the GDPR, and the system may be high-risk (Annex III, point 1). Emotion recognition in the workplace and in education has been prohibited since 2 February 2025, except for medical or safety reasons (Article 5(1)(f)).

Frequently asked questions

Does Article 50 of the EU AI Act already apply?

Yes. Article 50 has applied since 2 August 2026 and was not postponed by the Digital Omnibus (Regulation (EU) 2026/1744). Only the machine-readable marking of paragraph 2 has a transitional period: for systems placed on the market before 2 August 2026, it must be in place by 2 December 2026 (Article 111(4)).

Who must give the AI notice: the supplier or my organisation?

That depends on the paragraph. Paragraphs 1 and 2 rest with the provider, often the supplier of the chatbot or tool. Paragraphs 3 and 4 rest with the deployer, the organisation that uses the system. If you build a chatbot on someone else's model and offer it under your own name, you may be the provider yourself.

Does every AI-generated image need a label?

No. A deep fake, meaning AI-generated or manipulated image, audio or video content that would falsely appear to a person to be authentic or truthful, must be disclosed visibly or audibly by the deployer (paragraph 4). For other AI images you have no labelling obligation of your own under Article 50; the provider of the generator marks them in a machine-readable way (paragraph 2). Other law, such as the rules against misleading advertising, may still require a disclosure.

What is the fine for breaching Article 50?

Article 99(4)(g) sets a maximum of EUR 15 million or 3% of total worldwide annual turnover, whichever is higher; for SMEs, start-ups and small mid-cap enterprises the lower of the two applies. These are maximums that Member States must build into national law, and each Member State designates its own enforcement authorities. The Netherlands, for example, had not yet done so on 6 October 2026; the obligation itself still applies.

What is the code of practice on marking and labelling AI-generated content?

A voluntary code of 10 June 2026, with one section for providers and one for deployers. It covers paragraphs 2, 4 and 5. The Commission assesses whether adherence to the code is adequate for paragraphs 2 and 4 (Article 50(7)). The kit uses the code as a yardstick, even if you do not sign it.

May I adapt the notice texts and share them within my organisation?

Yes. You may use, adapt and share this template freely, including within your organisation, provided the credit 'Source: Praxikon' with the link to this template stays in place.

Use and credit

You may use, adapt and share this template freely, including within your organisation, provided the credit 'Source: Praxikon' with the link to this template stays in place.

How to cite this template: Source: Praxikon, AI transparency notice template for Article 50 of the EU AI Act, version 2.0, as of 6 October 2026, https://www.praxikon.com/en/templates/ai-transparency-notice

This template is a tool, not legal advice for your situation. It reflects the law as of 6 October 2026. Legislation, guidance and supervisory practice may change after that date. Using this template does not guarantee compliance: applying it in your organisation remains your own responsibility.

Praxikon is a trade name of Embed AI · Chamber of Commerce 90283597

Which obligation this belongs to