If you use Claude, the AI chatbot made by Anthropic, the text it writes for you is about to carry a hidden signature. The company has confirmed it will weave an invisible, machine-readable watermark directly into the words its models generate, so software can later flag the text as AI-made. You will not see it, and it does not change what Claude says. Anthropic says it has little choice: the mark is how it meets a new European transparency law that took effect on 2 August.
Why people are upset
The problem, for some users, is that the mark is baked into the text itself, so it survives copy and paste and can flag AI use long after the fact. Reddit lit up. One user, quoted by TechCrunch, likened the mark to a digital tattoo on the forehead for the student who used Claude to reorganise a paragraph or the journalist who summarised a long transcript. Others called the move unethical and disgusting, arguing that they, not the chatbot, did the real work. A counter-camp shrugged: the point is not credit, they said, but being able to tell when content is AI-made, given the risks that carries.
What the mark actually proves
Anthropic's own position is narrower than the panic. It says a detected mark is not fully conclusive, can appear even when a person supplied the ideas and only used Claude to polish, and can be scrubbed by heavy editing, paraphrasing or translation. The marking is applied at the model level and covers text across its products, from the chatbot to its coding and developer tools, and it also attaches signed provenance data to generated images and files under an open industry standard called C2PA.
Almost lost in the noise was a second move the same week: Anthropic turned its Claude in Chrome extension into a full Claude Cowork session that saves your chats and lets a task started in a browser tab finish in the desktop, web or mobile apps. It is live now for Max and Team subscribers and reaching others over the coming weeks, though it stays off by default for enterprise customers and does not yet run on other browsers or on mobile. One week, two messages: proving what AI made, and letting it do more.
