Cardano founder Charles Hoskinson has launched a free tool called Anthropies that removes Anthropic's watermarks from Claude-generated text, code, and images. The move raises questions about ownership and the legal implications of AI-generated content.
TL;DR
- Charles Hoskinson releases free tool to remove Anthropic's Claude watermarks, raising ownership questions.
- Anthropies uses a non-origin rewrite method to strip watermarks, with limitations when using Claude or Gemini.
- Hoskinson argues that Anthropic's terms could imply users don't own AI-generated content if they breach terms.
What happened
Charles Hoskinson, founder of Cardano, has released a free tool called Anthropies that removes Anthropic's watermarks from Claude-generated content. The tool was published on GitHub on Sunday, five days after Anthropic enabled invisible watermarking across all Claude products. Anthropies is licensed under Apache 2.0, a permissive open-source license.
Anthropies targets three layers of watermarking: git trailers, C2PA credentials on images, and text. The tool uses a non-origin rewrite method for text, routing it through a model other than Claude to avoid reapplying the watermark. However, this method has limitations and cannot be used with Claude or Gemini models.
Why it matters
The release of Anthropies raises important questions about the ownership of AI-generated content. Hoskinson argues that Anthropic's terms could imply that users do not own the content if they breach the terms of service. This could have significant implications for developers and startups using AI models to generate content.
The tool also highlights the ongoing debate about the transparency and control of AI-generated content. As AI models become more integrated into creative and professional workflows, the question of who owns the output becomes increasingly important. This could have implications for the competitive landscape and the future of AI-generated content.
Key facts
- Anthropies is a free tool released by Charles Hoskinson to remove Anthropic's watermarks from Claude-generated content.
- The tool targets three layers of watermarking: git trailers, C2PA credentials on images, and text.
- Anthropies uses a non-origin rewrite method for text, routing it through a model other than Claude to avoid reapplying the watermark.
- The tool is licensed under Apache 2.0, a permissive open-source license.
- Hoskinson argues that Anthropic's terms could imply that users do not own the content if they breach the terms of service.
- The release of Anthropies raises important questions about the ownership of AI-generated content and the transparency of AI models.
Context
The release of Anthropies comes at a time when the use of AI models in creative and professional workflows is increasing. As AI models become more integrated into these workflows, the question of who owns the output becomes increasingly important. This could have implications for the competitive landscape and the future of AI-generated content.
The debate about the ownership of AI-generated content is not new. As AI models become more advanced, the question of who owns the output becomes more complex. This could have implications for the legal and ethical frameworks that govern the use of AI models.