Transparency Templates for EU AI Act Art. 50: Ready to Use
Updated: June 2026 — Art. 50 in force from 2 August 2026
Introduction: The Transparency Obligation from August 2026
Art. 50 of the EU AI Act creates transparency requirements for AI systems that interact with people or generate content. These obligations take effect on 2 August 2026 — two months from now. Organisations that have not yet prepared their disclosures need to act.
This guide provides ready-to-use templates for the three most common Art. 50 disclosure scenarios, with adaptation guidance for different contexts.
The Three Art. 50 Obligations
Obligation 1: Chatbot Disclosure (Art. 50.1)
AI systems designed to interact directly with natural persons must inform those persons that they are interacting with an AI — not a human. This applies unless the AI nature is "evident from context."
When "evident from context" applies: A clearly branded AI assistant called "AIBot" on a website designed around AI interaction likely does not require an explicit disclosure. A conversational agent that could be mistaken for a human — especially in voice applications — requires explicit disclosure.
Obligation 2: AI-Generated Content (Art. 50.2)
Content produced by AI systems that is not clearly artistic or entertainment content must be labelled as AI-generated if it could be mistaken for authentic human-created content.
Obligation 3: Deepfake Disclosure (Art. 50.4)
AI-generated or AI-manipulated images, audio, or video depicting real or realistic-looking people, places, or events must carry a clear disclosure that the content has been artificially generated or manipulated.
Ready-to-Use Templates
Template 1: Chatbot Disclosure
Short version (for chat interface banners):
"You are chatting with an AI assistant. Responses are generated automatically and may not be 100% accurate. For important matters, please contact our team directly."
Full version (for terms of service or help text):
"This chat function is powered by artificial intelligence. The assistant can answer many questions, but is not a human and may make errors. Responses do not constitute legal, medical, or financial advice. For complex matters or when in doubt, please contact our [team/support]. [Organisation name] is responsible for this AI assistant."
Template 2: AI-Generated Written Content
For blog posts, reports, or other content substantially generated by AI:
"This content was (partly) generated with the assistance of AI tools and reviewed by [Organisation name]'s editorial team."
For full AI-generated documents:
"This document was generated by an AI system. The content has been reviewed but may contain errors. [Organisation name] takes responsibility for the accuracy of this document."
Template 3: Deepfake / AI-Generated Visual Content
For AI-generated images in marketing materials:
"Image created with AI"
For AI-manipulated or synthetic video:
"This video contains AI-generated content. The depicted persons/scenes may not be real."
Implementation Checklist for August 2026
- Identify all chatbots and AI conversational interfaces on your website, app, or customer service platform
- Implement chatbot disclosure messages in each interface
- Audit your content production process — identify where AI-generated content is used and label it
- Review any visual or video content for AI-generated elements requiring disclosure
- Update your content policy and brief your marketing and communications team
- Set a deadline of 31 July 2026 for full implementation
FAQ
Q: Does Art. 50 apply to internal AI tools (not customer-facing)?
A: Art. 50 applies to AI systems that "interact with natural persons" — which can include employees using internal AI tools. However, in the employment context where AI use is well-known and disclosed through other channels, the practical disclosure requirements may be lighter.
Q: Our AI chatbot is clearly named "AIBot" with a robot icon. Do we still need an explicit disclosure?
A: A clearly branded AI assistant is the strongest case for the "evident from context" exception. However, conservative practice — and easiest to defend in a supervisory audit — is to include an explicit disclosure regardless.