Anthropic recently decided to add watermarks to Claude’s outputs—an invisible code embedded in text edited or generated by the chatbot to identify content produced by artificial intelligence.
IT Home understands that Anthropic introduced this new policy to comply with the Transparency Code requirements of the European Union’s Artificial Intelligence Act (EU AI Act). The rules now require technology companies to label content generated or edited by artificial intelligence, and those labels must be detectable by computer systems. European regulators may be pleased, but some AI users are clearly unhappy.
On Reddit, it is easy to see users’ growing dissatisfaction, although other users on the site do not entirely agree with these views. One particularly heated post came from a user named visionode, whose account was created less than three weeks ago. Visionode argued that the new watermarking system was a harsh conspiracy designed to turn innocent chatbot users around the world into victims.
Visionode’s core argument seems to be that experienced Claude users can hide their use of AI by rewriting the content or processing Claude’s output through other AI services, while ordinary Claude users are not so lucky.
“Who will get caught? You. The student who used Claude to help reorganize a passage; the journalist who asked AI to summarize a 200-page interview transcript; the writer who hit a creative block and asked AI for synonyms. Once these people finish, they’ll all have a digital tattoo on their foreheads.”
These examples are not particularly apt. A journalist who asks AI to summarize a 200-page interview transcript would not be affected simply because the summary contains a watermark—unless they copied and pasted the AI-generated summary into their own article verbatim.
Similarly, if a student asks Claude to “reorganize a passage” and then directly copies and pastes Claude’s output into a paper, that practice is also inconsistent with academic ethics.
Other Reddit users did not particularly support the poster’s anger. One user simply commented, “Look at this guy.” Another told the original poster to “take a deep breath.”
Visionode was not the only person to complain. Another dissatisfied user called the watermarks “unethical” and “disgusting,” arguing that they had completed most of the work while using Claude. In their view, the chatbot was merely a “tool” that helped them do the hard work.
“I provided the instructions, background information, decisions, and countless revisions. Claude was just a tool. If Claude starts watermarking the code or anything else it generates, what exactly is it taking credit for?” the user asked.
Other users again pushed back against the critic. “It isn’t taking credit,” one user responded. “This is about being able to identify AI-generated content, because AI-generated content can pose risks in a variety of situations.”
Another user mocked them: “Bro, you can’t even roast Claude without having Claude write it for you.”
Some critics did not adopt the “victim” narrative, instead questioning Anthropic’s new policy from a more nuanced perspective.
For example, one user argued that watermarking content generated by an AI trained by extensively absorbing other people’s work reflects a broader double standard.
“I think this is a very sinister direction,” the user said. “I don’t use Claude to write anything, but considering how many cutting-edge AI models are trained by acquiring other people’s work as training data, having AI watermark your work carries a frightening irony in itself.”
Overall, however, users still seem inclined to support the watermarking mechanism, viewing it as a reasonable way to track algorithmically generated content.
“There is absolutely no good reason why this should not be a good thing,” one user in another discussion said. “The only reason you would not want such a mechanism is that you want to lie to other people.”