The Commission’s transparency FAQ (digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act) splits the work into four areas: people must know they are talking to an AI; generative systems must machine-mark synthetic audio, image, video and text; emotion-recognition and biometric-categorisation uses must be disclosed; and deployers who publish deepfakes — content resembling real people, places or events that would appear authentic — must put a visible or audible label at first exposure. AI-generated text on public-interest matters is labelled unless it had substantive human review or editorial control, not a spell-check.
Provider and deployer are different jobs. Providers of generative systems owe machine-readable marking, with a grace period to 2 December 2026 for systems already on the market before 2 August 2026, and watermark-detection interoperability running to 2 February 2027. Deployers who publish the deepfake owe the label the viewer sees. Artistic, creative, satirical or fictional works may label in a way that does not spoil the work. Standard assistive editing is exempt. Content generated and published before 2 August 2026 is not retroactive; generated before and published after is covered.
Three official EU icons exist for the visible layer: fully AI-generated, partially AI-modified, and a basic icon meant to sit with a short text label or an interactive second layer (digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content). Fines can reach €15 million or 3% of worldwide turnover. A platform AI toggle is not automatically this duty. The copy-paste labels and placement notes live in the disclosure template; this page is the name of the rule.
In practice
- Decide whether you are the provider (you built the generator) or the deployer (you published the clip). The duty is not the same.
- If a real person, place or event would look authentic, label at first exposure — first frame, not the description field.
- Public-interest text needs a label unless a human actually edited the claims, not just the commas.
The mistake to avoid
Treating TikTok’s or YouTube’s AI label as Article 50 compliance. It may help. It does not replace the legal test, which follows the audience and the content, not the app.
Go deeper
Article 50 disclosure template
Paste-ready Article 50 labels for video, image, audio and text, the three official EU icons, and where they go on TikTok, YouTube and Meta.
Related terms
Synthetic media disclosure
Synthetic media disclosure is the broader legal obligation to identify AI-generated or altered content as such — increasingly a binding requirement rather than a platform preference, with the EU AI Act's Article 50 as the clearest current example.
AI content label
AI content label meaning: a platform or creator tag that marks a post as made or altered with AI. Separate from paid-endorsement disclosure.
Content credentials
Content credentials are a cryptographically signed record of who or what made a file and what was done to it, attached by a hash computed over its exact bytes — which is why the record breaks the moment those bytes change, rather than degrading gracefully.
FTC endorsement disclosure
FTC endorsement disclosure meaning: say a material brand connection clearly where viewers cannot miss it, for human or AI talent.
Spark Ads
Spark Ads meaning: TikTok budget behind an existing post (yours or a creator's) so paid reach keeps the real likes, comments, and shares.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync, in your browser or on your phone.
Free on the web and on iPhone, one account.