Some Claude users are canceling their paid subscriptions after Anthropic introduced an invisible watermark designed to identify text generated by supported versions of its artificial-intelligence chatbot.
Anthropic began applying the “imperceptible watermark” to text produced by Claude models launched on or after August 2, according to reports. The marker is intended to remain associated with the text when it is copied and pasted, and may survive certain forms of editing. The company has also introduced provenance metadata for some AI-generated files, including images.businessinsider+1
The move is part of Anthropic’s effort to improve transparency around AI-generated content and comply with provisions linked to the European Union’s AI Act. Although the regulatory requirements are European, Anthropic’s watermarking policy is reportedly being applied globally rather than only to users in the EU.
Users raise privacy and professional concerns
The policy has triggered criticism among some writers, developers and other Claude users. Several users said they were concerned that clients, employers or readers could identify material created with Claude, even if the output had been substantially edited by a human.
One developer cited concerns that an AI label attached to code could create difficulties with clients. Other users objected to the absence of a visible opt-out and said they did not want their work to carry a persistent signal indicating that it had been generated or processed by an AI system.
Some paying customers have reportedly posted about canceling Claude Pro and Claude Max subscriptions. A professional quoted by Business Insider said he canceled a Claude Max plan, which costs at least $100 per month, because of concerns about the watermarking policy.
The backlash has also spread across social media and online forums, where users have discussed moving to alternative AI services or using locally hosted models that provide greater control over generated content. However, such posts are anecdotal and do not establish the scale of the cancellations.
Anthropic disputes cancellation trend
Anthropic has said it has not seen an increase in cancellations since announcing the watermarking feature. The company’s position suggests that while some customers have left in protest, there is no confirmed evidence of a large-scale subscriber exodus.
The distinction is important because publicly visible cancellations may represent only a small fraction of Claude’s overall user base. Neither Anthropic nor independent analysts have released figures showing how many subscriptions were canceled specifically because of the watermark.
The company says the watermark is intended to improve content provenance, not to reduce the quality of Claude’s responses or change their meaning. The system is designed to be difficult for ordinary users to detect, while allowing technical systems to identify the content as AI-generated.
Debate over AI transparency
Supporters of the policy say provenance tools could help distinguish AI-generated material from human work, reduce misinformation and strengthen accountability as generative AI becomes more widespread. They argue that invisible markers could be less intrusive than visible labels or mandatory disclosures attached to every piece of content.
Critics, however, question whether AI watermarks can reliably withstand extensive rewriting or whether they could unfairly stigmatize people who use AI as an editing or coding assistant. They also fear that customers could lose control over how AI-assisted work is presented to others.
For now, reports point to a limited but vocal group of Claude users canceling subscriptions over the change. A broader impact on Anthropic’s customer base remains unverified. The controversy nevertheless highlights a growing tension in the AI industry: companies are under pressure to make generated content traceable, while users want privacy, flexibility and control over how their work is identified.
