The EU AI Act's Transparency Rule Is Already Here. Most Bots Are Breaking It.
Since August 2, chatbots in the EU must say they're AI and deepfakes must be labeled. The high-risk bits were delayed. The honesty bit wasn't.
The EU AI Act's transparency obligations took force on August 2, 2026. The practical parts: any chatbot must disclose it's AI, and any AI-generated or manipulated content needs a label. These aren't future deadlines; they're live, and a lot of shipped products are silently non-compliant.
What got delayed was the heavy stuff. Regulation (EU) 2026/1744 pushed high-risk obligations to December 2027 for standalone systems and August 2028 for AI embedded in products. So the conformity assessments and human-oversight logs can wait, but the 'tell the user it's a bot' rule cannot.
For a site running AI comparisons or a chatbot, the fix is small: a line of disclosure and a watermark or label on generated media. The cost of ignoring it is a coordinated inspection, because the EU AI Office launched those in September and named missing documentation as a top finding.
The split is the story: Brussels made honesty mandatory now and capability-certification later. Builders should treat the disclosure as done-deal and stop shipping bots that pretend to be people, because that part of the law is already biting.