EU Mandates Labels for Realistic AI Images Starting Sunday

Beginning Sunday, the EU's AI Act mandates that technology firms must label AI-generated content that mimics individuals or real-life events, such as deepfakes, or risk substantial penalties. These regulations target content produced for professional use and automated informational text, while artistic, satirical, fictional, or personal content is exempt. Companies may implement watermarks or alternative detection methods. Google has endorsed the EU's AI transparency code, and Meta has indicated similar intentions. Both expressed apprehensions regarding the complexity of regulations and the need for adaptive technical solutions. Major platforms like Meta, which introduced its "AI Info" label in 2024, and TikTok, which has required labels for several years, likely have detection mechanisms established. Platforms are accountable for content created outside the EU that reaches European users if detection fails. Current tools, such as the C2PA protocol and SynthID watermark, have limitations.