AI Labelling
AI labelling refers to the obligation to make AI-generated content and AI interactions recognisable as such. It is the practical core of the Transparency Obligation under Article 50 of the EU AI Act and becomes binding on 2 August 2026. It mainly concerns chatbots, AI-generated text and images, and deepfakes on the websites of companies of any size.
What must be labelled?
Article 50 EU AI Act distinguishes three cases: first, chatbots and AI assistants must be recognisable – users must know before interacting that they are communicating with an AI and not a human (para. 1). Second, deepfakes – artificially created or altered image, audio and video content – must always be marked as artificially created (para. 4). Third, AI-generated text must be disclosed when it informs the public on matters of public interest and has not undergone human editorial responsibility.
From when and how is it labelled?
Visible labelling applies from 2 August 2026. For the additionally required machine-readable marking of AI-generated content (watermarks, metadata) under paragraph 2, the provider of the generative system is primarily responsible; for generative AI already on the market before the deadline, a transition period until 2 December 2026 applies here. For the typical website this means: a clear notice on the chat window and a label on AI content.
AI labelling in practice at BTECH
For AI features in client projects – such as an AI Chatbot or an AI-assisted feature – the transparency notice is built into the component from the start rather than retrofitted. That is the cheaper and legally sound route: a labelling notice already embedded in the code costs nothing in development.