AI glossary · Law and data protection

AI labelling

Also called: AI disclosure · Transparency obligations · Art. 50 EU AI Act

What is AI labelling?

Article 50 of the EU AI Act is about transparency: people should know when they talk to an AI and when content is generated. Duties fall partly on the provider of an AI system and partly on the deployer using it at work.

Two cases are typical for small businesses: a chatbot or AI phone assistant that talks to customers, and AI images in advertising.

How does it work?

The key paragraphs in plain words:

  • Para. 1, AI in conversation: people interacting with an AI system are told so, unless it is obvious. Formally this is the provider's duty, but the notice appears on your website or phone line.
  • Para. 2, marking: providers of systems that generate text, images, audio or video mark the output in a machine-readable way. Systems on the market before 2 August 2026 have until 2 December 2026.
  • Para. 4, deepfakes: if you use AI images, audio or video that resemble real people, places or events and would falsely appear authentic, you disclose that they are artificial.
  • Para. 5, when and how: at the first contact at the latest, clearly and accessibly.

Why does it matter for your business?

Because the duties have applied since 2 August 2026, and customers lose trust when they find out later that they talked to a machine.

An AI image that looks like a real photo of your business or a real job site can be a deepfake under the regulation, and even where it is not, it raises false expectations. This is an overview, not legal advice.

How does OsAI use it?

In the OsAI phone demo, “You are talking to an AI, not a human.” appears before the call, and every example call opens with the AI notice, as do assistants built for businesses.

OsAI labels AI-generated images as such, for example in an exposé film with a sample property. The free EU AI Act check gives a first assessment of your case.

Frequently asked questions

Do I have to label AI-generated text on my website?

By its wording, Art. 50 does not require labels on ordinary marketing copy. Para. 4 covers text informing the public on matters of public interest, and even then not if a person reviewed it and holds editorial responsibility. Check AI text before publishing anyway.

Does my chatbot have to say it is an AI?

Yes, unless it is obvious. Art. 50(1) requires that people learn they are dealing with an AI system, at the first contact at the latest. The simplest way is a clear opening line such as “You are chatting with an AI assistant.”

Do I have to label AI images in advertising?

If it resembles real people, places or events and would falsely appear authentic, it is a deepfake and you disclose that it is generated. For clearly artistic or fictional work, a discreet notice is enough. When in doubt, label it.

Sources

Last updated: