Claude Users Cancel Subscriptions Over AI Text Watermarking

Anthropic is facing customer pushback after implementing an "imperceptible watermark" in text generated by some versions of its Claude AI assistant. The feature, which went live globally on August 2, embeds a persistent marker that travels with content when copied and pasted, designed to distinguish AI-generated material from human writing.

Since the announcement Monday, dozens of users have publicly canceled their Claude subscriptions on X, with several confirming their decisions to Business Insider. The company maintains it has not observed a cancellation trend, though the backlash continues to spread.

Why it matters

The watermark controversy highlights a fundamental tension in enterprise AI adoption: companies need transparency around AI use, but professionals worry about how invisible markers might affect their work product, client relationships, and compliance obligations. As AI tools become embedded in workflows from coding to documentation, questions about attribution and control are moving from theoretical to immediate.

Professional concerns drive cancellations

Arturo Villarroya, who works in public policy, canceled his Claude Max subscription—priced at $100 monthly—on Wednesday after becoming uncomfortable with the watermarking plan. He uses Claude extensively for coding apps and tools, plus research and corrections at work.

Villarroya expressed concern that watermarks could persist even when Claude only translated text, checked spelling, or lightly edited original writing. In academic and professional environments with strict AI policies, he said, this could create problems.

Vladislav Rajtmajer, a freelance developer in the Czech Republic, cited similar worries when canceling his Max subscription Tuesday. He uses AI for code reviews and translations, and fears an AI label on client-shipped code could trigger questions about authorship or contract penalties.

Richard Echols, an AI consultant in Georgia, canceled Wednesday after concluding that even editing his own original work through Claude would add the watermark. "Even if you edit your own work that you wrote, they add the watermark anyway," Echols said.

Anthropic's position

Anthropic states the watermark indicates Claude processed text, not necessarily that it authored the content. The company says watermarking doesn't change meaning, quality, or readability of responses, and that marks can survive some editing.

The feature was implemented to comply with the European Union's AI Act, according to Anthropic. The company plans to release a free API allowing users and third parties to check text for Claude's watermark themselves.

Broader watermarking landscape

Anthropic isn't alone in watermarking AI output. Google uses its SynthID technology for AI-generated content, while OpenAI applies SynthID to supported images and audio. Elon Musk's X platform adds "Made with AI" tags to content it identifies as AI-generated or manipulated.

Proponents argue watermarks improve transparency as AI-generated content proliferates online, and could prevent AI systems from training on AI-generated material—a phenomenon that can degrade system performance.

Santiago Alessandro Rivera Martínez, a software engineer and digital agency founder, said the watermark reinforced his concerns about Claude dependency, prompting his cancellation 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.

Anthropic reported 300,000 business customers as of September 2025, a figure expected to have grown substantially since then, according to Business Insider, which first reported these details.