July 22, 2026. On August 2, 2026, the transparency obligations in Article 50 of the EU AI Act become enforceable. The rules govern how businesses disclose AI to the people they interact with, and they reach any company that puts AI generated content or AI chat in front of customers in the EU, wherever that company is based. The official summary is published by the European Commission, and law firm Sidley has a practical readiness guide.
What Article 50 requires
- Tell people when they are talking to AI. Systems that interact directly with users, such as chatbots and voice agents, must make clear that the person is dealing with an AI, unless it is already obvious.
- Mark AI generated content. Synthetic audio, image, video, and text must be labeled and machine readable as artificially generated or manipulated.
- Disclose deepfakes and public interest text. Deployers that produce deepfake image, audio, or video must disclose it, and AI generated text published to inform the public on matters of public interest must be disclosed.
One nuance keeps this from being a fire drill. Under the AI Omnibus provisional agreement from May 2026, generative AI systems already on the market before August 2 have until December 2, 2026 to meet the machine readable marking requirement in Article 50(2). The disclosure duties themselves still apply on August 2. The EU AI Office also published a draft Code of Practice on transparency in December 2025 to guide implementation.
What it means for operators
You do not need a legal department to get most of the way there. Three moves cover the common cases. First, add a visible AI disclosure to every customer facing chatbot and voice agent. Second, keep provenance on any AI generated marketing asset, using Content Credentials or your platform's labeling, so the machine readable marker travels with the file. Third, keep a simple record of where AI touches customer content, so a disclosure request becomes a lookup rather than an investigation. We bake these controls in from day one, both when we ship AI automation and when clients bring in an AI automation agency to retrofit an existing stack. Getting disclosure right is cheap when it is designed in, and expensive when it is bolted on after a complaint.
Frequently Asked Questions
It applies to providers and deployers of certain AI systems that interact with people or generate content, including companies based outside the EU when they serve EU users. If your chatbot, voice agent, or AI generated media reaches customers in the EU, the rules reach you.
The disclosure obligations apply from August 2, 2026. Under the May 2026 AI Omnibus agreement, generative AI systems already on the market before that date have until December 2, 2026 to meet the machine readable marking requirement in Article 50(2).
Yes. Synthetic audio, image, video, and text must be marked and detectable as artificially generated or manipulated. Deepfakes carry an added duty for the deployer to disclose that the content was AI generated or altered.
Disclose AI in chatbots and voice agents, keep provenance such as Content Credentials on AI generated assets so the marker travels with the file, and log where AI touches customer facing content so disclosure is a quick lookup.