Regulation July 20, 2026

The AI Act's transparency rules take effect on August 2: the labelling era for AI content

Article 50 of the EU AI Act enters into application on August 2, 2026: interactive AI systems must disclose themselves, deepfakes must be clearly marked, and AI-generated content must carry machine-readable labels. Generative systems already on the market were granted a watermarking grace period until December 2, 2026. Transparency violations carry fines of up to €15 million or 3 percent of global turnover.

A band of navy content blocks each receiving a small emerald stamp as they cross a vertical emerald threshold line

The EU AI Act’s broadest wave of application begins on August 2, 2026. Article 50’s transparency obligations require AI systems that interact with users to disclose themselves, deepfake content to be clearly marked, and the outputs of generative models to be labelled in machine-readable form. A compromise reached in May 2026 partially eased the timeline: generative systems already on the market have until December 2, 2026 to comply with watermarking, while the compliance date for stand-alone high-risk systems moved to December 2027.

For advertising and marketing, this has stopped being a legal footnote: for every campaign reaching EU consumers, identifying AI-generated content, marking it and auditing suppliers is now an operational process. The bill for a transparency violation can reach €15 million or 3 percent of global turnover. In an industry whose production is shifting to AI, labelling discipline is as much an infrastructure question as a creative one — who stamps what, and at which step?