Glossary Updates6 new terms added to the glossaries · October 11, 2026, 13:38 CEST
AI TechDocKnowledge

Glossary · EU AI Act: general-purpose AI and transparency

Deepfake (AI Act)

Also known as: Deep fake, Deepfake disclosure, Article 3(60)

German: Deepfake

In the EU AI Act, a deep fake is AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful (Art. 3(60)); deployers who generate or manipulate such content must disclose that it is artificial (Art. 50(4)).

  • AI regulation
  • EU
  • USA
  • China

In one sentence

Deepfake under the AI Act (Art. 3(60)): AI-made image, audio or video that falsely appears real; deployers must disclose it from August 2, 2026.

Example

A marketing team uses an AI tool to show a real service engineer speaking a language he does not speak in a training video; the video must disclose that it was artificially generated or manipulated.

How it applies

  • Definition: The definition covers image, audio and video, not text. It applies to content resembling real persons, objects, places, entities or events, and only where it would falsely appear authentic or truthful. Regulation (EU) 2026/1744 did not change Article 3(60).
  • Disclosure by deployers: Deployers of an AI system that generates or manipulates deep fake content must disclose that it was artificially generated or manipulated (Art. 50(4)). For evidently artistic, creative, satirical or fictional works, the duty is limited to disclosing the existence of such content in a way that does not hamper the work. An exception applies where the use is authorized by law to detect, prevent, investigate or prosecute criminal offenses.
  • Marking by providers: Separately, providers of generative AI systems must mark synthetic output in a machine-readable format (Art. 50(2)). Article 50 applies from August 2, 2026; for systems placed on the market before that date, the Digital Omnibus on AI gives providers until December 2, 2026 to meet the marking duty.
  • Timing and form: The information must be clear and distinguishable and given at the latest at the time of the first exposure (Art. 50(5)).
  • Technical documentation: Training and marketing videos that show real people or real equipment doing what they did not do fall under the definition when they could be taken as real. Illustrations that are evidently drawn or rendered usually do not. Disclosure informs viewers; it does not make the content accurate, and AI-generated visuals in instructions still have to show the product correctly.

Compared with the USA, Canada and China

China regulates deepfakes most broadly: the deep synthesis provisions and the AI-generated content labeling measures require labels for synthetic content in general. The USA has no general federal labeling duty; the federal TAKE IT DOWN Act (Public Law 119-12, signed May 19, 2025) targets non-consensual intimate images including AI-generated “digital forgeries”, with platform removal duties enforceable since May 19, 2026, and several states regulate election or other deepfakes. Canada has no general federal deepfake disclosure duty and relies on its voluntary code. See transparency obligations (AI Act) for the other duties of Article 50.

External references