Since 2 August, AI in Europe has to say it is AI
The EU AI Act's transparency rules are now in force and the Commission's AI Office can enforce them. Chatbots must identify themselves, deepfakes must be labelled, and AI generated content needs machine readable marks. Fines reach 15 million euros or 3 percent of global turnover.
A quiet but real change landed in Europe on 2 August. From that date, the European Commission’s AI Office, together with national authorities, can actually enforce the AI Act, and a new set of transparency rules started to apply. If you use AI tools in Europe, or build anything with them, this is the week the rulebook got teeth.
The core of it is short enough to say in three sentences. Chatbots and other interactive AI systems have to tell you that you are dealing with software and not a person. Deepfakes, meaning images, video or audio that were generated or edited with AI, have to be labelled. And AI generated or altered content has to carry machine readable marks, which means an invisible signal inside the file that other software can check automatically, so a platform or a fact checker can detect it without a human squinting at pixels.
The Commission also published a first list of more than 180 organisations that have signed a Code of Practice on transparency of AI generated content, which is the voluntary document that spells out how to meet the rules in practice. Penalties for ignoring the obligations run up to 15 million euros or 3 percent of worldwide annual turnover, whichever is higher. There is one important piece of slack: generative systems that were already on the market before 2 August have until 2 December 2026 to meet the machine readable marking requirement, under the AI Omnibus agreement from May.
What is actually going on here
Regulation usually arrives in two stages. First the law exists on paper, then someone gets the power to knock on a door. The AI Act was passed back in 2024, but until now the enforcement machinery for general purpose AI models, meaning the big general models that get used for everything, was not switched on. Now the AI Office can request technical documentation, evaluate models itself, demand corrective measures and issue fines. It also opened a complaints tool and a whistleblower channel, which is a signal that it expects tips from inside companies rather than only from its own testing.
A fair caveat: labelling is not the same as solving anything. Machine readable marks can be stripped, and a “this is AI” notice at the top of a chat window is easy to stop noticing after a week. The rules reduce accidental deception more than deliberate deception.
What this means for you: if you are just an AI user, expect more small disclosure banners and more labelled images in your feeds over the coming months, and take the absence of a label as weak evidence rather than proof. If you run a website, a shop or a small business that uses a chatbot or publishes AI generated images, you are a deployer under this law, and the honest first step is a plain sentence telling visitors when they are talking to a bot. That costs nothing and covers most of what is being asked.
Sources
Google reshuffles its AI leadership, and four of its most famous engineers walk out the door
Demis Hassabis moves from Google DeepMind CEO to Alphabet chief scientist, while Jeff Dean, Sanjay Ghemawat, Oriol Vinyals and Quoc Le leave to start a company called Discovery Loop. The flagship Gemini model is still not out.
