People Are Rushing to Find Ways to Remove Claude’s AI Text Watermark

Earlier this month, Anthropic started placing an imperceptible label on all text produced using its Claude AI model to serve as an indicator that the material was generated with AI. The folks who don’t want it outed that they are using AI for work are now spending their hours looking for a way to scrub their outputs of that marker.

As Business Insider notes, the term “AI watermark remover” has started trending on Google , jumping 60% in search interest week over week, and an open-source GitHub project called Watermarks Remover has amassed more than 15,000 “stars” and thousands of forks with the promise to scrub any output of indicators that it came from AI. Sabrina Ramonov, an AI educator, created her own web-based tool for washing text of residual model markers.

Why the panic, exactly? Well, according to the creator of the Watermarks Remover tool, it’s a matter of privacy and attribution. “I am all for content attribution,” Guillaume Meyer, the creator of the popular GitHub project, told Business Insider. “I am against the watermarking technique, and that’s a very significant distinction.”

It’s probably fair to say that an AI watermark will create a false binary in many people’s minds. If some content carries the invisible marker, they’ll simply assume the entirety of that material was created with Claude, when it’s possible that only a small portion of it was AI-generated. And it’s not like people aren’t already out here trying to hit people with AI accusations. One only has to look at how popular Pangram and other so-called “AI detection” tools have gotten to see how weaponized something like this could be (albeit, most of the time the stakes are pretty low and amount to little more than “AI shaming”).

Likewise, there is concern that by slapping this marker on content, Anthropic and other companies that institute a similar watermark may be laying the groundwork for claiming ownership over certain material. Anthropic seems to have thought about that, and mentioned in a blog post that “A watermark only helps test whether Claude might have produced or processed the content. It doesn’t say anything about ownership or authorship and doesn’t change a user’s rights under our terms.” Of course, that’s only true until it isn’t. A blog post isn’t binding.

The only thing that is for sure is that the cat-and-mouse game is likely to continue. Anthropic is issuing its watermark to comply with the EU AI Act , which requires AI system providers to create methods of “marking” outputs. But currently, there’s no rule keeping a person from removing that watermark. And folks have been seeking ways to scrub AI content of these indicators since the start—there are dozens of tools out there claiming to provide the same washing techniques for AI-generated images and videos, too.

Here’s a question: how many of those tools do you think were made with AI?