EU AI transparency rules take effect, require disclosure of chatbots and deepfakes

The European Union's AI Act transparency obligations took effect on August 2nd, establishing a new regulatory baseline for how companies must disclose AI use to the public. The rules emerged from the EU's landmark AI Act and are designed to address a growing problem: as generative and interactive AI systems proliferate, distinguishing AI interactions and AI-generated content from authentic human content has become increasingly difficult.

The rules impose different obligations on two categories of actors. Providers—companies that develop and market AI systems—must design their systems to explicitly notify users when they are interacting with AI rather than a human, "unless this is obvious." They must also embed machine-readable marks into synthetic audio, images, video, and text so that artificially generated or manipulated content can be automatically detected. Deployers—platforms and services that use AI systems—face complementary requirements: they must label any AI-generated or manipulated image, audio, video, or deepfake content that is designed to look real. Some companies, including Meta and SpaceX AI, are classified as both providers and deployers, subject to both sets of obligations.

To ease compliance, the European Commission developed a set of AI disclosure labels that tech platforms can adopt. These labels mirror designs already introduced by TikTok, Instagram, and Facebook. The Commission provided these examples to prevent platforms from having to design their own labeling schemes, though while the EU labels are optional, the Commission emphasized that the labeling requirements "are not." The underlying rationale is straightforward: "People should know when they are interacting with AI or exposed to AI-generated content. This will help them make informed decisions, calibrate their trust and reliance on AI and avoid misinformation or deception."

Noncompliance carries significant penalties. Companies that fail to follow the new transparency rules risk fines of up to €15 million (about $17.2 million), or 3 percent of their global annual turnover—whichever is higher. However, the timeline differs for existing and new systems. AI models and services that launched before August 2nd have a four-month grace period, until December 2nd, to bring themselves into compliance. New AI systems introduced on or after August 2nd must comply immediately.