Anthropic's new AI watermark prompts several Claude subscribers to quit

Anthropic disclosed that some versions of its Claude assistant will now embed a hidden watermark in the text they produce, a tag that can survive copy-paste and limited editing. The company says the marker does not alter meaning or readability, but users fear it may linger in content they later modify. Since the policy took effect on August 2, dozens of subscribers have announced cancellations on X, uneasy about the watermark showing up in code, research or client documents.

One outlet spoke with four former Claude Max customers—including Arturo Villarroya, Vladislav Rajtmajer, Richard Echols and Stevan Bogosavljević—who each highlighted different professional risks and have migrated to other AI services. Anthropic maintains that the watermark merely indicates Claude processed the text, not that it authored it, and plans to release an API for detecting the mark.