EU AI Act transparency rules take effect: fines up to €15M for non-compliance

On August 2nd, the European Union activated the transparency pillar of its AI Act, forcing companies to clearly flag when people interact with AI systems and when content is AI-generated or manipulated. The rules distinguish between providers—those developing and marketing AI systems—and deployers, such as platforms like TikTok or Meta, that integrate those systems into their services. Providers must design AI systems that explicitly notify users they are dealing with a machine unless it's obvious, and embed machine-readable markers into synthetic audio, video, images, and text so the content can be automatically detected. Deployers, meanwhile, must visibly label AI-generated or manipulated deepfake content designed to look authentic.

The European Commission also released a standardized set of AI disclosure icons that platforms can adopt to stay consistent, complementing similar labels already used by TikTok, Instagram, and Facebook. While these EU-designed icons are optional, the underlying labeling requirements are not. Non-compliance carries penalties of up to €15 million (about $17.2 million) or 3% of a company's global annual turnover, whichever is higher. New AI systems face immediate enforcement, but models and services already on the market before August 2nd have a grace period until December 2nd to comply. The Commission says the rules aim to help people calibrate trust, make informed decisions, and avoid misinformation—critical as generative AI blurs the line between human and machine output.