Why Claude Users Are Canceling Over Anthropic's Watermark

Some Claude users have canceled their subscriptions after Anthropic said certain versions of the chatbot would embed an invisible watermark in their text output.

Anthropic said this week that the watermark will travel with content when copied and pasted and may survive some editing, helping to distinguish AI-generated material from human writing. The company said watermarking does not change the meaning, quality, or readability of Claude's responses.

Despite that, dozens of people posted on X since Monday that they had canceled their subscriptions, citing the watermark. Business Insider spoke with four users who said they had canceled. Some shared confirmation of their cancellations and said they disliked the prospect of a persistent marker in their work and were worried about their AI use showing up in client work, such as coding.

Anthropic told Business Insider it has not seen an uptick in cancellations since it announced the watermark.

Arturo Villarroya, who works in public policy, canceled his Claude Max subscription Wednesday after becoming uncomfortable with the watermarking plan, which has applied globally to text from supported Claude models since August 2. He said he uses Claude for coding, apps, sourcing, research, and corrections at work. He worried a watermark could remain even when the underlying writing was original and Claude was used only to translate, check spelling, fix citations, or lightly summarize a passage. The watermark could create problems in academic and professional settings with strict rules on AI use, he said. He is switching to Cursor and Grok for coding, and Grok Bot for work and personal organization.

Anthropic said in its blog that a detected watermark indicates Claude processed text, not that it was necessarily the original author. Marks can remain after Claude proofreads, translates, summarizes, or otherwise edits content.

Vladislav Rajtmajer, a freelance developer in the Czech Republic, canceled his Claude Max subscription on Tuesday, citing the watermark as the main reason. He uses AI for code reviews and translations and worried an AI label on code shipped to clients could raise questions about authorship or trigger contract penalties. The watermark was a "signal to look at Chinese AI providers and learn to work with other models," he said.

Richard Echols, an AI consultant in Georgia, said he canceled his Claude Max subscription on Wednesday, and the watermark was a major factor. He uses Claude to create documentation for businesses and worried the mark could appear when he had written the underlying material himself and used the chatbot only for edits. "Even if you edit your own work that you wrote, they add the watermark anyway," Echols said.

Stevan Bogosavljević, a senior software engineer in Montenegro, canceled his Claude Max subscription in July, citing declining service quality, limited access to Anthropic's top models, and stronger alternatives. The watermark further validated his decision to cancel. He said having code attributed to Claude would not be desirable because professionally shipped code carrying an invisible marker could raise questions about authorship, compliance, or client policies.

Not everyone is against watermarks. Some say they improve transparency as more online content is generated by artificial intelligence. Others say they could help prevent AI systems from being trained on a growing volume of AI-generated material, which can cause systems to degrade. Aadit Sheth, cofounder of executive communications firm The Narrative Company, said on X on Tuesday that audiences should be able to tell whether the words they read reflect a person's own thinking and that watermarks are one way to ensure that.

Anthropic is not the only AI lab to use watermarks. Google uses its SynthID technology to watermark AI-generated content, and OpenAI uses SynthID for supported images and audio. X also adds a "Made with AI" tag on content it determines to be AI-generated, though the tag appeared and later disappeared on outgoing White House press secretary Karoline Leavitt's resignation post.

Anthropic said it added the watermark to comply with the European Union's AI Act and plans to release a free application programming interface that will allow users and third parties to check text for Claude's watermark themselves.

Santiago Alessandro Rivera Martínez, a software engineer and founder of the digital agency Directa, said the watermark reinforced his concerns of becoming too dependent on Claude and led him to cancel his subscription this week. "When your workflow is built entirely around proprietary software, the provider can change the rules, features, restrictions, or conditions around that workflow later," he said. "I would rather have the AI layer of my work be portable and provider-agnostic."