A market for removing Claude’s new invisible AI watermarks has emerged, led by open-source projects, web tools, and commercial evasion services claiming to clean marked text. Anthropic says its watermarking is meant to satisfy EU AI Act transparency rules, but the actual watermark is hard to verify because no public detector has been released yet. #Claude #Anthropic #EUAIAct #watermarks-remover #StealthGPT #HumanWrites
Keypoints
- Tools to remove Claude watermarks appeared within days of Anthropic’s disclosure.
- watermarks-remover is one of the largest projects and now claims broad model support.
- Many tools only strip hidden characters or metadata, not the actual watermark.
- Anthropic says its watermark is woven into word choice and can be lost through heavy rewriting.
- Security concerns remain because these tools are being added directly into agent workflows.