EU AI Act: AI chatbots, cloned voices and deepfakes must now be disclosed
The transparency rules of the EU AI Act have applied since 2 August 2026. The Digital Omnibus pushed back the rules for high-risk AI, but the duty to tell people when they talk to an AI or see a deepfake stayed on schedule.
What applies since 2 August 2026
Article 50 of the AI Act sets transparency duties for three situations that matter to anyone using a digital twin or a cloned voice:
- Chatbots: people must be told that they are talking to an AI system, unless it is obvious from the context.
- Deepfakes: whoever publishes image, audio or video content that imitates real people, places or events and could pass as authentic must disclose that it was generated or manipulated by AI.
- Generative AI providers: the output of their systems must be marked in a machine-readable way, so it can be detected as AI-made.
What the Digital Omnibus moved, and what it left alone
Regulation (EU) 2026/1744, known as the Digital Omnibus on AI, postponed the obligations for high-risk AI systems: to December 2027 for the systems listed in Annex III and to August 2028 for AI built into products covered by Annex I. The disclosure duties for chatbots and deepfakes were not postponed.
The one transition left in Article 50 concerns machine-readable marking: generative AI systems that were already on the market before 2 August 2026 have until 2 December 2026 to comply. New systems had to comply from launch.
The European Commission published guidelines on Article 50 in July 2026, next to a code of practice on the transparency of AI-generated content with one part for providers (marking and detection) and one for deployers (labelling).
What it means if you use a digital twin or a voice clone
- Say it is an AI. A chat window that answers for you should make clear that visitors are talking to your digital twin.
- Label what you publish. A video narrated in your cloned voice, or showing a generated version of you, should carry a clear AI label where you post it. YouTube asks about altered or synthetic content during upload, and Instagram adds an AI info label.
- Clone only what is yours. Use your own voice and likeness, or have written consent from the person you clone.
- Keep a record of that consent and of where the content was published.
How TalkMie works with these rules
A TalkMie twin is presented as the digital twin of its owner, on its public page and in the chat window. A voice is cloned only with the owner's consent and only for the owner's own twin. Owners can also list topics their twin must not discuss, such as medical or legal advice. When you publish a TalkMie video narrated in your cloned voice, add the platform's AI label as described above.
See also: Voice cloning, explained and Denmark's plan to protect your face and voice.
This article is general information about the law, not legal advice. National enforcement differs between member states.
Questions
Do I have to label a video narrated in my own cloned voice?
If the video could pass as a real recording of you, it is a deepfake in the sense of Article 50 and should be disclosed as AI-generated where you publish it. Labelling it costs nothing and keeps you on the safe side.
Did the Digital Omnibus delay the deepfake rules?
No. It delayed the obligations for high-risk AI systems. The disclosure duties for chatbots and deepfakes have applied since 2 August 2026; only machine-readable marking for systems already on the market has a grace period until 2 December 2026.
Does content made before 2 August 2026 need a label?
According to the Commission, retroactive labelling of older deepfakes is not mandatory, but it is encouraged.