Labelling of synthetic content is a transparency obligation under which content created or modified by artificial intelligence must be identifiable. It covers machine-readable marking on the part of providers of generative systems, and a visible notice on content that depicts real people, objects, or events in a way that could be misleading. Exceptions apply to works that are manifestly artistic or satirical, where the label must not disrupt the work. For companies this is relevant to advertising, videos with avatars, synthetic voice, and edited photographs. Independently of this obligation, the rules on misleading advertising still apply – depicting a product in a way that does not correspond to reality is a breach even when it is labelled as generated. Internal rules should therefore cover who decides on labelling and how it is recorded which output was created by generation.
See also: AI Video Generator, Voice cloning and AI dubbing, Unfair commercial practices.
