Cardano founder releases tool to strip Claude’s new AI watermarks

Anthropic has begun embedding invisible watermarks into text generated by its Claude AI models, a move that has triggered user backlash and already prompted the release of a counter-tool designed to strip the markings.

How the Watermark Works

The watermarking system, which Anthropic detailed in a blog post on Friday, uses Google Alphabet Inc. ↘ 0.12% DeepMind's SynthID-Text approach to create a statistical pattern in Claude's word choices that is imperceptible to readers but detectable with the correct cryptographic key. When Claude generates text, it often faces "low-stakes choices" between equally suitable words — such as "overcast" versus "grey" — and the watermarking system steers those selections according to a secret key.techcrunch+2

Anthropic said the watermark "does not impact the quality of Claude's output" and has "negligible" effect on speed or cost. The company plans to release a watermark detection API, though details of its implementation remain forthcoming.ndtvprofit+3

The system applies globally to Claude models launched after August 2, 2026, the date the EU AI Act's transparency code took effect. Anthropic said it is applying the watermark worldwide because it does not yet have "a durable way to scope it by region." Light editing is unlikely to remove the watermark, though a complete rewrite replacing every word will.indiatoday+3

User Backlash and Subscription Cancellations

The announcement, first made on August 11, has provoked strong reactions. Business Insider News Corp ↗ 0.73% reported that "dozens" of users on X claimed to have canceled their Claude subscriptions. Gizmodo noted that math and AI influencer John Ennis posted a screenshot of his cancellation, citing the "ridiculous watermark idea." On Reddit, users characterized the move as everything from a "conspiracy against pure users" to a necessary step against deception.news.sbs+2

Anthropic told Business Insider that it has not seen an increase in cancellations since the announcement.gizmodo

A Counter-Tool Emerges

On Sunday, Cardano founder Charles Hoskinson published a free, open-source tool called "Anthropies" — a portmanteau of Anthropic and herpes — designed to remove the watermarks. Released under an Apache 2.0 license, the tool routes Claude-generated text through a different language model to rewrite it, destroying the statistical pattern in the process.firstpost+1

Hoskinson framed the release as a warning about ownership risks, arguing that the watermark and Claude's co-authorship attribution could create legal complications for users down the road. He also flagged the risk of false positives — an editor who runs human-written copy through Claude for a quick proofread could later have that work flagged as AI-generated.tech.yahoo+1

Anthropic has not publicly responded to the tool. The company noted in its blog post that other major AI developers have signed the same EU Code of Practice and will implement their own watermarking systems.mezha+1