AI Watermark Removers Emerge as Anthropic Rolls Out Labeling

Anthropic last week began rolling out an imperceptible watermark on text generated by certain Claude models. The company said the statistical pattern carried by word choice would travel with the text when copied or pasted, making AI-generated content easier to identify.

Not everyone welcomed the move. Some Claude users told Business Insider they canceled their subscriptions because they feared the label would follow work even if they used the AI only to proofread, translate, or summarize. Google Trends interest in “AI watermark remover” among U.S. users rose 60% week over week.

Developers responded quickly. Guillaume Meyer, a Paris-based tech entrepreneur and founder of Memo, built an open-source project called Watermarks Remover within days of Anthropic’s announcement. The tool strips hidden characters and metadata, then rewrites text to maintain its meaning, disrupting the word-choice patterns that carry the watermark, Meyer said. The first version took about five hours to build. The project has more than 14,000 stars on GitHub, a measure of popularity among developers. Meyer said it does not guarantee removal.

“I am all for content attribution,” Meyer told Business Insider. “I am against the watermarking technique, and that's a very significant distinction.” He argued that watermarking treats authorship as a binary question, tagging output the same way for full generation and light editing. Anthropic said in its blog the watermark is not intended to establish authorship.

Sabrina Ramonov, an AI educator for entrepreneurs, posted on X this week that she created a free, browser-based Watermark Remover that can “clean hidden AI marks” from text, PDFs, Word documents, web pages, images, and data files. Ansh Aneja, a Tokyo-based software developer, built a Claude-focused watermark remover the day Anthropic announced the feature. He later released a local open-source version called MarkScrub. In an X post last week, Aneja said the tool went from zero to 8,500 users in one day. Business Insider could not independently verify those figures.

These projects are early. Business Insider did not independently assess their effectiveness. Other companies, including Google and OpenAI, also use watermarks for images, and many tools already promise to remove them. Still, they illustrate a long-standing limit of the technique, said Thibaud Gloaguen, a researcher at ETH Zurich’s Secure, Reliable, and Intelligent Systems Lab. “There will always be ways to remove the watermark,” he told Business Insider, citing rewording as an example.

Legal questions remain unresolved. The European Union’s AI Act tells AI providers to add durable labels but does not clearly address third parties trying to remove them. A European Commission spokesperson told Business Insider that the bloc’s guidance requires providers’ marking systems to be resilient to common alterations and adversarial actions. The transparency code specifically lists removal, regeneration, copying, and modification as threats providers should assess. Those rules apply to AI providers. The AI Act does not expressly bar third parties from removing watermarks, said Dmitri Roussinov, a senior lecturer in Computer and Information Sciences at the University of Strathclyde. If a removal tool uses AI to regenerate text, however, its provider may have its own obligation under EU law to mark that newly generated output, he said.

Konrad Kollnig, an assistant professor at Maastricht University’s Law and Tech Lab, said neither the AI Act nor Anthropic’s terms appear to bar people from making or sharing removal tools. Users could face problems if they use one to misrepresent AI-generated work as human-made. Anthropic’s usage policy prohibits impersonating a human by presenting model results as human-generated.

That question could gain urgency when Anthropic releases its planned text-detection API alongside its next model. The company has not specified a date. Watermarks will apply to new models released on or after August 2. Anthropic also plans to roll it out to older models. “All these watermarking techniques have one big issue,” Kollnig said. “Whenever the watermarking detection tool is made public, anyone can check AI-generated contents – including text and videos – against that detector and can build tools to remove watermarks.”