Article 50 covers four categories: systems that interact with people, systems generating synthetic content, emotion recognition and biometric categorisation, and systems producing deepfakes. It applies regardless of whether the underlying system is classified as high-risk, which is the part most marketing teams missed.
The duty that lands on an advertiser is the deployer duty: disclose that content has been artificially generated or manipulated, clearly and distinguishably, at the latest at first exposure. A disclosure at the end of a thirty-second spot does not satisfy it and neither does a policy page.
It follows the output rather than the company. A UK or US brand serving a campaign to EU audiences is inside it, which for most paid social is the default rather than a decision. Penalties run to fifteen million euro or three per cent of worldwide annual turnover.
Does Article 50 apply to companies outside the EU?
It follows the output. Where a system’s output is used in the Union, the obligations attach regardless of where the brand, agency or studio is established.
What does the December 2026 grace period cover?
Only the machine-readable marking obligation in Article 50(2), and only for systems placed on the market before 2 August 2026. It is a provider concession about watermarking infrastructure, not a deferral of the deployer disclosure duty.
Is there an artistic exemption?
There is a reduced obligation for evidently artistic, creative, satirical or fictional work. It does not cover creator-format advertising, where the persuasive effect depends on the content not reading as fiction.