|

Cardano Founder Launches New Free Tool to Remove Anthropic’s AI Watermark

Cardano founder Charles Hoskinson has revealed a free software that removes Anthropic’s watermarks from Claude output. He named it Anthropies, a mix of Anthropic and herpes.

The challenge appeared on GitHub on Sunday, 5 days after Anthropic switched on invisible marking throughout each Claude product. Hoskinson shipped it beneath an Apache 2.0 license.

How the Anthropic Watermark Actually Works

Anthropic now marks Claude text on the mannequin degree. The rollout solutions the European Union AI Act transparency code, which took impact on Aug. 2.

The mark is just not a hidden character string. Instead, a secret key steers event sampling, a course of that breaks ties between phrases becoming equally nicely.

Anthropies. Source: Hoskinson Github

Hoskinson’s repository splits the issue into three layers. The “Co-Authored-By” git trailer strips out cleanly. C2PA credentials on photographs disappear after a re-encode.

Prose is the exhausting layer. The software subsequently routes textual content by means of a mannequin apart from Claude, a step the repository calls a non-origin rewrite.

That design carries a blunt limitation. Running the rewrite inside Claude or Gemini would stamp the mark straight again on, so the ability refuses.

The repository ships three modes. Clean strips trailers deterministically, humanize rewrites prose elsewhere, and orchestrate blocks the job when the host mannequin is Claude.

Code barely carries a sign in any respect. Syntax leaves nearly no room for substitution, so the watermark has nowhere to sit.

The Apache 2.0 license issues right here. It is a permissive open-source time period that lets anybody copy, modify, promote, or bundle the code, supplied they maintain the unique notices. It additionally grants a patent license, so Anthropic couldn’t block forks on patent grounds.

“The mark is the wording. There isn’t any separate payload to delete. Removing it means altering which phrases are there.”

Anthropies README

Diagram of the three Anthropic watermark layers. Source: Hoskinson Github

Hoskinson framed the discharge on X as a warning somewhat than a utility. He mentioned the watermark and the Claude co-author line might trigger points down the highway.

Roughly half the repository is just not code. Hoskinson assaults the project clause that assigns possession of output to customers, which applies “topic to your compliance with our Terms.”

He reads that phrase as a situation precedent, that means a hurdle customers should clear earlier than possession passes. A breach of phrases might subsequently imply possession by no means transferred.

The git trailer attracts separate hearth. A broadcast co-authorship line might later floor as proof of joint authorship, he argues. He additionally flags a threat of false positives. An editor who runs human copy by means of Claude for a fast proofread might journey a detector months later.

AI defaults now arrive unasked, as Twitch confirmed when it opted creators into training. Anthropic has stayed quiet to date. The firm is in the meantime making ready a listing above $2 trillion, and it has already withheld its strongest model from the general public.

Hoskinson, for his half, has spent 2026 selecting fights over technical credit score, together with a declare that Ethereum copied Cardano’s ledger design.

The repository counted 4 stars on Sunday morning. Adoption issues lower than the query it poses. Who owns a sentence as soon as the mannequin has signed it?

The put up Cardano Founder Launches New Free Tool to Remove Anthropic’s AI Watermark appeared first on BeInCrypto.

Similar Posts