Market for AI watermark removal tools emerges after Anthropic’s Claude update

As reported by Bleeping Computer, a market for tools designed to remove AI watermarks has rapidly emerged just days after Anthropic announced the implementation of invisible watermarks in content generated by its Claude AI model.

This development includes a GitHub project with over 4,500 stars, various new web tools, and existing AI detection evasion services. While claims of successful watermark removal are widespread, they remain unverified as Anthropic has not yet released its detection methods. Tools like Guillaume Meyer's "watermarks-remover" and others such as "claude-watermark-cleaner" and "remove-ai-watermarks" have appeared, with some claiming to support watermarks from OpenAI and Google's Gemini as well. These tools often focus on stripping metadata or hidden characters, which is a less complex task.

The actual removal of the AI watermark, embedded within the model's word choices, would require significant text rewriting, a process that some developers, like Meyer, admit is not yet functional in their tools. The emergence of these tools highlights a potential supply chain risk, especially as the market shows significant demand and verification remains challenging, particularly in light of the EU AI Act's transparency requirements.

Source: Bleeping Computer