Claude Watermark Cracked: Open‑Source Remover Gains 15K+ Stars
Anthropic added an invisible watermark to Claude's outputs, prompting the community to create the open‑source watermarks‑remover tool that strips three layers of hidden marks—from zero‑width characters to statistical sampling watermarks and file metadata—supporting multiple AI models and file formats, and it quickly amassed over 15,000 GitHub stars.
Claude watermark deployment
Anthropic announced that Claude models released after 2023‑08‑02 embed an invisible watermark in every generated text to satisfy EU regulatory requirements. The watermark is applied globally to all Claude outputs.
watermarks‑remover project
A GitHub repository https://github.com/guillaumemeyer/watermarks-remover was created within a week of the announcement and quickly attracted over 15,000 stars. The project provides two usage modes:
Agent skill callable via a chat‑bot interface.
Standalone Python HTTP service.
The tool can strip watermarks from multiple AI providers (Claude, Gemini, OpenAI) and from more than twenty file formats, including PNG, PDF, DOCX, and MP4.
Three‑layer removal architecture
Layer 1 – invisible characters : Detects and removes zero‑width spaces, unusual whitespace characters, and bidirectional control characters. These markers are explicit and can be eliminated deterministically with simple scripts.
Layer 2 – statistical sampling watermarks : Targets watermarks encoded in the distribution of words across a paragraph. Because no specific characters are present, the layer rewrites the text to disperse the statistical signal.
Layer 3 – provenance metadata : Strips C2PA provenance information from images and EXIF metadata from PDF and Office documents.
Implementation details
The core implementation relies exclusively on the Python standard library; no third‑party packages are imported. A single command launches a local HTTP server, allowing the service to be used without embedding the code in another project. The default local callback address does not invoke any vendor APIs, and files in unsupported formats are rejected outright to avoid unintended modifications.
Agent skill endpoints
Initially the Agent skill was named /remove-claude-marks and handled only Claude watermarks. After extending support to other AI providers, the endpoint was renamed /remove-ai-marks, while the original Claude‑specific command remains functional.
Usage constraints
The author notes that the tool should be applied only to content owned by the user or for which the user has explicit permission.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
IT Services Circle
Delivering cutting-edge internet insights and practical learning resources. We're a passionate and principled IT media platform.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.