EU now requires labels on AI chatbots and deepfake content

New transparency rules under the EU AI Act force companies to flag AI chats and mark synthetic media, with fines up to €15 million for skipping it.

What happened: On August 2nd, new transparency rules under the EU's AI Act kicked in. Companies that build or deploy AI systems in the EU now must tell people when they're talking to a bot rather than a human, and must mark content that AI has generated or altered — images, audio, video, and text — so it can be identified as synthetic. The European Commission also released a set of standard disclosure icons companies can use, similar to labels TikTok, Instagram, and Facebook already show on AI content.

Why it matters: Generative AI has made it harder to tell real from fake online, and the Commission says that erodes trust and opens the door to deception. These rules are meant to let people 'calibrate their trust' in what they see and hear. It's one of the first times a major government has forced disclosure, rather than leaving it to platforms' voluntary policies, across such a broad range of AI products at once.

How it works, plainly: The law splits responsibility two ways. 'Providers' — the companies that build AI models — must design systems to say upfront when someone is dealing with AI, unless that's obvious, and embed machine-readable marks in anything synthetic they output. 'Deployers' — the apps and platforms that use those models — must separately label realistic-looking deepfake images, audio, or video for their users. Some companies, like Meta, count as both and carry both duties.

The rollout: The rules are enforceable immediately for any AI system launched on or after August 2nd. Systems already on the market before that date get a grace period until December 2nd to comply. Companies that miss the deadline or ignore the rules risk fines of up to €15 million (about $17.2 million) or 3 percent of their global annual revenue, whichever is larger.