The transparency obligations are the rules in Article 50 of the EU AI Act on when people must be told that they are dealing with AI. They apply from 2 August 2026.
There are two kinds of obligation. Providers, who make the tools, must ensure that content their systems generate is marked in a machine-readable format, and that a chatbot is designed so users can see it is a machine. Organisations that use the tools must disclose deepfakes and AI-generated text on matters of public interest if the text is published without human editorial control and responsibility.
The requirements do not mean that all AI-assisted content must be labelled. A press release that an employee has written with help from AI, edited and stands behind does not trigger a disclosure obligation.
In practice
Review your touchpoints: chatbots on the website, automated replies in customer service, and synthetic voices or images of real people. That is where the obligations typically apply. Decide at the same time when you will disclose AI use yourselves, even where the law does not require it. That decision belongs in your editorial AI policy.
See also AI labelling and my guide to the EU AI Act.
The concepts in this glossary run through my AI courses and workshops, where we work with them on your own tasks.