watermarks-remover Strips AI Provenance Marks

AI-generated content is supposed to carry invisible watermarks that stay with it. A tool called watermarks-remover says it can strip those marks from work made with Claude, Gemini, and OpenAI models. It installs as an agent skill, not a standalone app.

What Happened

watermarks-remover works in layers. It removes invisible Unicode and bidi characters, rewrites text to break statistical watermarks, and cleans C2PA, EXIF, and XMP metadata from PNG, JPEG, SVG, PDF, DOCX, HTML, and Markdown files. The core scripts are built on the Python 3.10+ standard library. Soft-binding and audio watermarks are not supported.

Why It Matters

The tool targets provenance marks that AI companies use to label generated content, including C2PA metadata. That makes it possible to scrub the source of an image or text after it leaves the model. For anyone tracking AI governance or content provenance, this is a direct challenge to the idea that watermarks will hold.