AI Signal 444
Anthropic Watermarks Text Claude Processes or Generates, Sparking Subscription Cancellations
↗
Anthropic added watermarks to text that Claude generates or processes, including proofreading, translation, and editing, prompting users to cancel subscriptions over concerns that human-authored content touched by Claude could be flagged as AI-generated.
Engineers using Claude for code reviews, translations, or documentation edits risk having their own work marked as AI-processed when shipped to clients, potentially triggering contract penalties or authorship disputes. The watermark persists even when Claude only touches the edges of human-written content, and some users report it incorrectly claims full authorship of work it did not create.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Anthropic's watermark flags text Claude processed (edited, proofread, translated), not just text it generated, meaning human-written content edited by Claude carries the mark.
Multiple users canceled Claude Max subscriptions, citing concerns that the watermark could raise authorship questions or trigger contract penalties on client-delivered work.
Some users report Claude incorrectly claims full or co-authorship of content it only touched peripherally, with instances dating back to Sonnet 4.6.
THE CLUSTER