Cardano Founder Launches Tool to Remove Claude AI Watermarks

Cardano founder Charles Hoskinson has released a new free tool called Anthropies, aimed at removing invisible markers associated with content generated by Anthropic’s Claude AI models. The release comes shortly after Anthropic introduced new watermarking and attribution mechanisms designed to help identify AI-generated content.

The tool reportedly targets several types of markers, including signals embedded in generated text, C2PA metadata associated with images, and Git-related attribution information in code. These technologies are intended to provide greater transparency around AI-generated material, particularly as governments and technology companies increasingly focus on identifying synthetic content.

Hoskinson has questioned the implications of Anthropic’s approach, particularly the possibility that AI attribution could create complications for people who legitimately use AI-generated material. His tool therefore represents a direct challenge to the idea that AI-generated content should automatically carry persistent attribution information.

The development also highlights a growing technical and policy debate around AI provenance. Companies are increasingly exploring watermarks, metadata and other mechanisms to distinguish machine-generated content from human-created work. Such systems could help platforms and users understand where content originated, but they also raise questions about privacy, permanence, accuracy and who should control attribution.

The release is notable because it comes from a major figure in the cryptocurrency industry rather than an established AI company. It demonstrates how the debate over AI transparency is spreading beyond the traditional AI sector. As generative AI becomes increasingly common in writing, programming and image creation, the question of how—and whether—AI-generated material should be identifiable is likely to remain a major issue for developers, businesses and regulators.