Europe’s AI labeling and transparency rules are now in effect
The Verge Jess Weatherbed ● Covered by 59 sources
The EU's new AI transparency rules kicked in on August 2nd. Now companies have to tell you when you're chatting with a bot or looking at AI-made content.
Brussels just flipped another switch on its AI Act, and this one is about honesty rather than raw safety. As of August 2nd, companies operating in the EU have to make clear when someone is talking to a chatbot instead of a human, and when an image, video, or piece of text has been generated or manipulated by AI. No more guessing whether that eerily articulate customer service rep is a person or a language model wearing a friendly avatar.
The rule splits obligations between two groups: providers, who build and sell the AI systems, and deployers, who actually put those systems in front of users. A company that makes a chatbot has to design it so it can't pretend to be human. A company that deploys someone else's chatbot on its website has to make sure users know they're not texting a person. Some firms, like Meta and SpaceX, land in both buckets at once, since they build models and also run the products people actually use.
To take some of the guesswork out of compliance, the European Commission rolled out a set of standardized labels and icons that companies can slap on AI-generated content instead of inventing their own disclosure system from scratch. It's a small but telling move — Brussels clearly wants consistency across the bloc rather than a patchwork of vague watermarks and fine-print disclaimers that vary by platform.
This is part of the AI Act's staggered rollout, which has been arriving in phases since the law passed last year, with harsher rules for high-risk AI systems still to come. Transparency is the easier lift compared to what's ahead — verifying model safety, auditing training data, policing systemic risk from the biggest foundation models. But even this modest step forces a shift in how AI companies build products for European users: disclosure isn't optional anymore, it's baked into the design requirements.
Whether this actually changes user behavior is another matter. People already skim past cookie banners and terms-of-service pop-ups without a second thought, and an AI label could easily join that pile of ignored notifications. Still, having the requirement on the books gives regulators something concrete to enforce, and gives researchers and journalists a paper trail when companies quietly cut corners.
My take
I like that the EU is forcing disclosure by default rather than hoping companies volunteer it, because left alone, most of them wouldn't bother. But standardized icons only matter if regulators actually check whether they're being used honestly, and Brussels has a spotty track record of enforcement keeping pace with its own ambitious rulebooks. Transparency theater is easy; transparency with teeth is the part that's still unproven.
Read more about this at: The Verge
