Charles Hoskinson, the founder of the Cardano blockchain, has published a free open-source release that removes the watermarks Anthropic now attaches to Claude output. He calls it Anthropies.
The release lands five days after Anthropic switched on invisible marking across every Claude product. The code appeared on GitHub under an Apache 2.0 license, so anyone can copy, modify, or sell it.
Anthropic's mark is not a hidden character. A secret key steers tournament sampling, the step that breaks ties between words that fit equally well, so the signal lives in word choice rather than a payload you can delete.
Hoskinson's repository attacks the problem in three layers. The "Co-Authored-By" git trailer strips out cleanly. C2PA credentials on images disappear after a re-encode. Prose is the hard case, so the code routes text through a model other than Claude, a step it calls a non-origin rewrite, and it refuses to run that rewrite inside Claude or Gemini because they would simply stamp the mark back on.
He packaged the release with three modes: Clean strips the trailers deterministically, humanize rewrites prose elsewhere, and orchestrate blocks the job when the host model is Claude.
Hoskinson framed the project as a warning more than a utility. His legal argument centers on Anthropic's assignment clause, which grants users ownership of output "subject to your compliance with our Terms." He reads that phrasing as a condition precedent, so a breach could mean ownership never transferred. He also argues the co-authorship line could later surface as evidence of joint authorship, and that an editor who runs human copy through Claude for a quick proofread could trip a detector months later.
Anthropic has not responded. The company is preparing a public listing above two trillion dollars and has already withheld its strongest model from the public.
The repository had collected four stars by Sunday morning. Adoption is not really the point. The question it poses is simpler: once a model has signed a sentence, who owns it?
