Deepfake Legal Framework Takes Effect in the EU: AI-Generated Content Must Carry Legally Binding Labels

On July 1, 2028, the EU's Deepfake and Synthetic Media Act (DSMA) formally took effect. This is the world's first legislation specifically targeting AI-generated content. Its core requirement: all AI-generated or AI-assisted content circulated within the EU—text, images, audio, and video—must carry legally binding digital labels.

The DSMA's labeling requirements operate on three tiers. The first is metadata labeling: all AI-generated content file metadata must include the generative model name, generation time, and generator identity. The second is visible labeling: AI-generated images and videos must display an "AI Generated" watermark in the corner; AI-generated text must note at the beginning that it was produced by AI. The third is verifiable labeling: through EU-certified blockchain evidence systems, each piece of AI content receives a tamper-proof digital fingerprint.

Violations face severe penalties: first-time fines up to 2% of global annual revenue, repeat violations up to 6%. For individual users, fines range from 5,000 to 50,000 euros.

EU Commission President von der Leyen said: "The DSMA's goal isn't to restrict AI use but to ensure people know what they're seeing. In an era where AI can generate convincing content, labeling rights are citizens' fundamental right to know."

Meta, Google, and TikTok have all announced they will integrate DSMA labeling features in their EU product versions. But implementation faces technical challenges—particularly for text content, where no reliable technology can distinguish human writing from AI writing.

China's Cyberspace Administration indicated on July 3 that it is researching similar AI content labeling regulations, with a draft expected by end of 2028.