
EU AI Act watermarking mandate takes effect August 2026
EU Enforcement Deadline for AI-Generated Content Detection
The European Union has formally adopted Regulation (EU) 2024/1689, the EU AI Act, establishing binding transparency obligations for generative AI system providers. Article 50 of the regulation mandates that providers mark outputs generated by AI systems in a machine-readable format so that content is detectable as artificially generated or manipulated. These obligations apply from August 2, 2026 onwards.
The requirement covers synthetic audio, image, video, and text content produced by qualifying AI systems. The machine-readable marking standard ensures that detection tools—whether automated systems or human-operated software—can identify synthetic material consistently across platforms and applications.
Grandfathering Rule Extends Compliance Window
On May 7, 2026, EU legislative bodies reached a political agreement on proposed amendments to the AI Act that introduced a grandfathering provision. Under this rule, generative AI systems already placed on the market or put into service before August 2, 2026 are granted a delayed compliance deadline. Instead of meeting the watermarking requirement on August 2, these existing systems must comply by December 2, 2026—a four-month grace period.
This staged implementation recognizes the practical burden on vendors with deployed systems while maintaining a firm enforcement schedule. Systems placed on the market or deployed on or after August 2, 2026, face immediate compliance obligations upon launch.


