Anthropic's watermarking policy sparks debate: human text edited by AI may be flagged as AI-generated

max_paperclips · x · 2026-08-11

A user points out that under Anthropic's new watermarking policy, not only will text written by Claude be marked as AI, but if a user uploads their own text and asks Claude to copy-edit it, the resulting text will also be watermarked as AI. This policy raises concerns about the attribution of AI-assisted content.

Related event: Anthropic's Watermark Policy Sparks Controversy(2 posts)→

Original post →

More from Safety

Safety channel →