EU Mandates AI Content Watermarking Starting August 2

As of August 2, 2026, an EU AI Act requirement takes effect: companies must attach a digital marker — a watermark — to all AI-generated content that looks like the real thing. The law covers images, audio recordings, and text. Exceptions are made for private chats and clearly artistic or satirical works.

Violations carry fines of up to 3% of a company's global annual revenue. Systems already operating in the European market get an additional four months to comply. Companies may use the standard labels proposed by the EU or develop their own, provided they meet the technical requirements.

The law aims to protect consumers and combat disinformation. The Member of the European Parliament shepherding the legislation called it a tool for "preserving democracy and the authenticity of facts online." The requirement primarily targets content that, without a label, would be nearly indistinguishable from human-made — deepfakes, synthetic voices, and AI-written news.

This is the first large-scale legal instrument against deepfakes and disinformation backed by real financial penalties. Global platforms with European users — from Google to TikTok — are now required to build out watermarking infrastructure. It sets a precedent that other regulators are certain to follow: the UK and Canada are already watching closely.