Some Claude users are mad that Anthropic’s new watermarks will catch them using it at their jobs, classes

Anthropic has started watermarking Claude’s outputs, slipping invisible code into the chatbot’s text so it can be identified as AI-generated. The move is tied to the EU AI Act’s Transparency Code, which requires tech companies to mark content that’s been generated or edited in a way computers can detect.

That has not gone down well with everyone. Reddit has turned into the complaint desk, with some users arguing the system will catch ordinary people using Claude for work, class, or writing help. One especially heated post came from a user named visionode, whose account was only three weeks old, and who framed the change as a kind of global trap for unsuspecting chatbot users.

The examples in that post are easy to recognize: a student fixing a paragraph, a journalist summarizing a long transcript, a writer looking for synonyms. But the real problem with those examples is less the watermark and more what happens next. If someone copies Claude’s text straight into an essay or article, the issue isn’t a hidden code. It’s plagiarism with extra steps.

Other Redditors didn’t seem moved by the outrage. One told the poster to take a deep breath. Another mocked the complaint for apparently being written with Claude. There were broader objections too, including the argument that it’s ironic for an AI system trained on other people’s work to mark its own output as if that were the scandal.

Still, the dominant reaction in the thread was approval. Some users said there’s no good reason not to watermark AI output, and that the only people who’d object are the ones who want to mislead others about where the text came from. That’s probably the real line here: transparency versus plausible deniability. Anthropic picked transparency, and a few users immediately noticed how much they’d been relying on the other thing.