Article 50 EU AI Act: the 5-point checklist for businesses
From 2 August 2026, chatbots must disclose that they are AI, and AI-generated content must be labelled. This affects not just big tech but every company with a chatbot on its website or AI-generated content in its marketing. This checklist shows what to do in practice.
Last updated: July 2026 — not legal advice
Article 50 applies regardless of risk class and has no size threshold. It covers four situations:
Chatbots, voice bots and AI assistants must be designed so users realise they are talking to an AI — unless it is obvious from the context.
Providers of generative systems must mark outputs (text, image, audio, video) as AI-generated in a machine-readable way, e.g. via metadata or watermarks.
Anyone deploying such systems must inform the people exposed to them.
Deployers must visibly label deepfakes — and AI-generated or -manipulated text published to inform the public on matters of public interest.
By 2 August 2026 you should have worked through these points:
Generative AI systems already on the market before 2 August 2026 have until 2 December 2026 to implement machine-readable marking. All other obligations — in particular chatbot disclosure and visible deepfake labelling — apply from 2 August 2026 without deferral.
Breaches of the transparency obligations can be fined with up to EUR 15 million or 3% of global annual turnover. In Germany, the Bundesnetzagentur becomes the central AI market-surveillance authority.
Yes. Article 50 has no size threshold — what matters is whether an AI system interacts directly with people or generates content. A small web shop with a chatbot is affected too.
No. The information must reach the person at the moment of interaction — for a chatbot, that means directly in the conversation, at the latest at first interaction, clearly and unmissably.
Breaches of the transparency obligations can be fined with up to EUR 15 million or 3% of global annual turnover. In Germany, the Bundesnetzagentur will take over central market surveillance.
Yes. The disclosure duty for chatbots hits the provider via the system design, but as the deployer you must make sure the disclosure actually reaches users in your deployment context — and the deepfake and text labelling duties fall directly on the deployer.
We audit your chatbots and AI content for Article 50 conformity, clarify your role as provider or deployer, and implement the labelling technically — pragmatic and without the hype.
Or start with our structured AI potential and compliance analysis.
Go to the AI readiness analysis