Anthropic's new AI watermark sparks subscription cancellations
Users express discomfort with persistent markers in Claude-generated content
1What is the watermark?
Anthropic announced it has added an 'imperceptible watermark' to text generated by its Claude AI assistant. This watermark can persist even when content is copied, pasted, or lightly edited. The company claims the watermark doesn't alter the meaning, quality, or readability of responses. However, this practice has sparked significant backlash, particularly among professional users who rely on Claude for sensitive work.
2User backlash
Following the watermark announcement, dozens of users took to X (formerly Twitter) to share they had canceled their subscriptions. Four users speaking with Business Insider showed cancellation confirmations, expressing discomfort with the watermark. Developers using AI for client work like coding are particularly concerned it might raise authorship questions or violate contract terms regarding AI usage in deliverables.
3Professional concerns
Public policy professional Arturo Villarroya canceled his $100/month Claude Max subscription, worried the watermark might persist even when Claude is only used for minor tasks like fixing academic citations or summarizing passages. Similarly, a freelance developer from the Czech Republic feared AI labels on client code could trigger contractual penalties, stating this fundamentally changes the value proposition of the service for professionals.
4Migration to alternatives
Canceling users report switching to alternatives like Cursor, Grok 4.6, and Grok Bot. A Montenegrin software engineer said the watermark decision was the final straw after already considering cancellation due to declining service quality and limited access to top models. Some users are now exploring Chinese AI providers, seeing this as an opportunity to diversify their AI toolset beyond Western offerings.
5Anthropic's defense
Anthropic clarified the watermark indicates Claude processed text but doesn't necessarily mean it was the original author. The company stated the watermark complies with the EU AI Act and promised a free API for users to check texts themselves. However, these explanations haven't alleviated concerns for many professional users who feel the implementation doesn't account for legitimate use cases where AI assists rather than replaces human work.
6Industry practices
Anthropic isn't alone in watermarking AI content. Google uses its SynthID technology to mark AI-generated material, while OpenAI employs similar systems for supported images and audio. Elon Musk's X platform adds 'Made with AI' labels to content it identifies as AI-generated or manipulated, though enforcement appears inconsistent - as seen when a label briefly appeared then disappeared on a White House resignation post this week.
7Pro-watermark arguments
Not all voices oppose watermarking. Advocates argue it increases transparency as AI-generated content proliferates online. Aadit Sheth, co-founder of The Narrative Company, stated audiences deserve to know if words reflect human thinking. Some experts also suggest watermarks could prevent AI systems from training on synthetic data, which might degrade model performance over time through a phenomenon called 'model collapse' in machine learning circles.
8Broader implications
With Anthropic reportedly serving 300,000 business customers as of September 2025 (a number likely grown since), the long-term impact of watermark backlash remains uncertain. However, professional users' sensitivity highlights the delicate balance AI companies must strike between transparency and usability. This incident will likely fuel ongoing global debates about AI ethics, responsible use policies, and how to maintain trust in human-AI collaboration environments where attribution matters.
Original article: Business Insider — Anthropic'in yeni AI filigranı abonelik iptallerine yol açtı