Deepfakes and AI avatars.
Reviewed
Short answer
A deepfake is AI-generated or manipulated image, audio or video that resembles existing persons, objects, places, entities or events and would falsely appear authentic or truthful, and anyone using AI professionally to make one must disclose it clearly at the first exposure. Realistic AI avatars count: the Commission’s guidelines list a synthetic avatar of a company CEO and an AI-generated celebrity influencer in an advert as deepfakes.
Official sources: AI Act, Article 3(60) · Commission guidelines on Article 50, section 6.1
The four criteria
| Criterion | What it means | Example that fails it |
|---|---|---|
| Resemblance | A high, appreciable similarity to the subject, judged case by case; it need not be identical | — |
| Existing | Resembles someone or something that exists, could plausibly exist or could have existed | Humans flying without aids, dragons, elephants driving cars |
| Persons, objects, places, entities or events | Includes digital replicas, realistic AI-generated human avatars or personas, and traits such as voice or behaviour | — |
| Falsely appears authentic or truthful | Judged on the content as a whole, its message, the context and the expected audience; no intent to deceive is required | Background scenes or special effects in standard film production |
AI avatars, digital humans and synthetic influencers
A realistic AI-generated person is a “person” for this test even if no such person exists, because it could plausibly exist. The guidelines’ deepfake examples include a realistic synthetic avatar of a company CEO congratulating employees, an AI-generated celebrity influencer in an advert, and a realistic synthetic influencer testing a sponsored product. If the avatar talks with people in real time, its provider also has to tell them they are interacting with an AI under Article 50(1).
Art, satire and fiction: lighter, not exempt
| Gets the lighter disclosure | Does not |
|---|---|
| Films or trailers with de-aged actors or digital replicas of dead actors, in cinemas or on streaming services | A teleshopping-style video with simulated customers selling a product |
| AI music resembling the individual style of existing artists | AI images of celebrities in activities that never happened, with no satirical or fictional purpose |
| A manipulated image of a politician in a scene clearly meant to criticise their policies with humour | A realistic synthetic influencer testing a sponsored product |
| Gaming imagery with deepfake simulations of real people | Realistic AI video of Holocaust scenes shared on public social media |
The work must be evidently artistic, creative, satirical or fictional to its audience, and the guidelines read this strictly: if the content is ambiguous, or mixes an informative and a creative character, the informative character prevails and the normal label applies. The disclosure can then be placed so it does not spoil the work, for example in the credits or the description (Code of Practice, Section 2, Commitment 3).
Banned, not just labelled, from 2 December 2026
Regulation (EU) 2026/1744 added two prohibitions to Article 5: AI systems that generate or manipulate realistic images, videos, audio or similar material of an identifiable person’s intimate parts or of them engaged in sexually explicit activities without their freely given, specific, informed, unambiguous and explicit consent (point (ba)), and AI systems that generate child sexual abuse material (point (bb)). For providers, the ban covers systems whose intended purpose this is, or whose design makes it a foreseeable and reproducible outcome without adequate safeguards; for users, it covers using a system for that purpose.
Platform rules for deepfakes
| Platform | Rule | Source |
|---|---|---|
| YouTube | Disclosure required for realistic content that, for example, makes it look as if a real person was arrested, shows a public figure stealing, or makes someone appear to give advice they did not give. | YouTube Help |
| Facebook, Instagram, Threads | Labels AI content and may add a more prominent label when digitally created or altered images, video or audio carry a particularly high risk of materially deceiving the public on a matter of importance. | Meta |
| TikTok | Requires labels on AI-generated or significantly edited content showing realistic-looking scenes or people, and does not allow content that is misleading about matters of public importance or harmful to individuals, even with a label. | TikTok |
| Google Ads (election ads) | Advertisers must disclose election ads with synthetic or digitally altered content that inauthentically depicts real or realistic-looking people or events. | Google Ads policy |
Questions
- Are deepfakes illegal in the EU?
- Not as such under the AI Act: professional deployers must disclose them. Two kinds are prohibited from 2 December 2026, non-consensual sexual deepfakes of identifiable people and child sexual abuse material, and other laws, such as data protection or intellectual property law, can apply to any deepfake.
- Does a parody need a label?
- Yes, but a lighter one, if it is evidently satirical to its audience: the disclosure must not hamper the work. The guidelines give a humorous manipulated image of a politician criticising their policies as an example, and they read “evidently” strictly.
Sources
- AI Act, Article 3: DefinitionsEuropean Commission, AI Act Service Desk (AI Act Explorer)Consolidated text as at 27 July 2026 (includes the Digital Omnibus, Regulation (EU) 2026/1744), shown in the Commission’s AI Act Explorer.read
- AI Act, Article 50: Transparency obligationsEuropean Commission, AI Act Service Desk (AI Act Explorer)Consolidated text as at 27 July 2026 (includes the Digital Omnibus, Regulation (EU) 2026/1744), shown in the Commission’s AI Act Explorer.read
- AI Act, Article 5: Prohibited AI practicesEuropean Commission, AI Act Service Desk (AI Act Explorer)Consolidated text as at 27 July 2026 (includes the Digital Omnibus, Regulation (EU) 2026/1744), shown in the Commission’s AI Act Explorer.read
- Guidelines on the transparency obligations under Article 50 of the AI Act (C(2026) 5054, 20 July 2026)European Commissionread
- Transparency obligations under Article 50 of the AI Act — questions and answersEuropean Commissionread
- Code of Practice on Transparency of AI-Generated Content (final text, PDF)European Commissionread
- Disclosing use of GenAI contentYouTube Helpread
- Our Approach to Labeling AI-Generated Content and Manipulated Media (updated 1 July 2024)Meta Newsroomread
- Community Guidelines: Integrity and Authenticity — Edited Media and AI-Generated Content (AIGC)TikTokread
- Political content — disclosure requirements for synthetic or digitally altered contentGoogle Advertising Policies Helpread
Last reviewed by Volkan Çatak, Founder, Proofcan.
Not legal advice. This page explains official texts in plain words and links each one; check the sources, or a lawyer, before relying on it for a decision.