The European Union published a Code of Practice on June 10 under Article 50 of the EU AI Act, laying out specific rules for how platforms and content creators must label AI-generated images, audio, video, and text — including deepfakes — before users encounter them.
The rules cover two groups: AI providers, which build the underlying models, and AI deployers, which use those models to generate and publish content. For deployers, the requirements are concrete. Deepfakes — defined as AI-generated or manipulated media that “would falsely appear to a person to be authentic” — must carry a visible disclosure label at the moment of first exposure, not buried in credits or metadata. AI-generated text published to inform the public without human editorial oversight falls under the same obligation.
The EU has produced three standardised icons that platforms can use free of charge and without attribution: “AI GENERATED” for fully synthetic content, “AI MODIFIED” for partially manipulated material, and a minimal “AI” icon that can be paired with a custom interactive layer. The icons were tested across EU member states for noticeability and clarity before being finalised, with user testing finding that labels including a clear text component outperformed icon-only versions. Equivalent custom labels are permitted, but must preserve the capitalised “AI” acronym as the primary visual element and follow the same design proportions.
Placement rules vary by content type. For images and video, the label goes in an unobstructed position — the Code suggests the top-right corner — and for video it must reappear at the start and after any ad breaks. For published text, the disclosure sits above or near the headline. For audio-only content, a spoken disclaimer in plain language is required at the start, with reminders repeated throughout for live or long-form audio; when a screen is also present — a car dashboard or smartphone — a visual icon is required alongside the spoken notice. The Code also mandates accessibility compliance: audio descriptions for visual labels, tactile or haptic cues for audio content, and screen-reader compatibility.
Artistic and satirical content that is clearly recognisable as such sits outside the mandatory disclosure scope, as does content approved through human editorial review — the rules target autonomous AI publishing pipelines. Platforms covered by the Digital Services Act are the primary near-term enforcement targets as the AI Act continues its phased rollout, which began in 2024.
For anyone who scrolls social media, watches YouTube, or reads news online in the EU, this means AI-generated content should start showing visible labels at the top of posts, in video frames, and before audio clips — wherever platforms implement the Code. For the platforms themselves, the placement and design specifications are detailed enough to require real engineering work across content pipelines, not just a policy update.
About Dignited Staff
This account is managed by the in-house team at Dignited, a collective of passionate tech writers, editors, and enthusiasts dedicated to bringing you the latest insights, reviews, and news on consumer technology. For inquiries or feedback, feel free to reach out to us at [email protected].
Discover more from Dignited
Subscribe to get the latest posts sent to your email.
