The recent decision by Anthropic to implement watermarks on text generated by its AI assistant, Claude, is causing some users to reconsider their subscriptions. This week, the AI lab announced that certain versions of Claude will now include an "imperceptible watermark" within the text output, allowing it to remain detectable when content is copied, pasted, and even edited. This initiative aims to distinguish AI-generated writing from that created by humans.
Anthropic maintains that the watermark does not alter the meaning, quality, or readability of the responses generated by Claude. Nevertheless, numerous users have taken to social media, particularly X, to announce their cancellation of Claude subscriptions, largely due to this new feature.
Business Insider interviewed several former subscribers who expressed their concerns. For instance, Arturo Villarroya, a public policy professional, canceled his Claude Max subscription—priced at $100 monthly—after feeling uneasy about the watermarking system. Since August 2, this watermark has been universally applied to text from supported models of Claude. Villarroya has relied on Claude for various tasks, including application coding, sourcing information, and conducting research. He worries that even original text could inadvertently carry the watermark, especially in contexts with strict guidelines regarding AI utilization.
Another user, Vladislav Rajtmajer, a freelance developer in the Czech Republic, echoed these sentiments, stating that he was apprehensive about how the watermark could affect client perceptions of authorship and could even lead to contractual issues. He mentioned he is now considering other AI options, including Chinese providers.
Richard Echols, an AI consultant based in Georgia, also cited the watermark as a significant reason for his subscription cancellation, as he uses Claude for business documentation. He expressed concern that the watermark would appear even when he was the primary author, undermining the integrity of his work.
In addition to user pushback, Stevan Bogosavljević, a senior software engineer in Montenegro, canceled his subscription due to perceived declines in service quality. He found the watermark to be an additional reason to seek alternatives, emphasizing that he would not want generated code to be marked in a way that raises questions about compliance or authorship.
While there is considerable negative feedback regarding the watermark, there are also proponents who argue it promotes transparency in an age where AI-generated content is becoming increasingly prevalent. Some, like Aadit Sheth, cofounder of The Narrative Company, believe watermarks can help clarify whether a piece of writing reflects individual thinking.
Anthropic is not alone in its approach to watermarking; other companies, such as Google and OpenAI, have implemented similar technologies to identify AI-generated content. Additionally, Elon Musk's platform X has introduced a "Made with AI" label for identifiable AI-generated material.
Anthropic claims that the watermarking decision aligns with compliance efforts for the European Union's AI Act. The company also plans to introduce a free API that will allow others to check for the presence of a Claude watermark in the text.
Santiago Alessandro Rivera Martínez, a software engineer and founder of Directa, recently decided to cancel his subscription, voicing concerns about being overly reliant on any one proprietary software. He emphasized the importance of having a flexible, provider-independent approach to incorporating AI into his workflow.
