Anthropic's Claude watermarking sparks backlash from users who fear detection

Anthropic has begun watermarking Claude’s outputs, embedding invisible code into generated text that allows computer systems to identify it as AI-produced. The move is designed to satisfy the EU AI Act’s Transparency Code, which requires tech companies to label AI-generated or AI-edited content in a machine-detectable way. While regulators may welcome the change, a segment of Claude users is pushing back. On Reddit, one user called the watermarking a “draconian conspiracy,” arguing that students who “reorganize a paragraph” or journalists who summarize long transcripts will be unfairly exposed. Others dismissed the outrage, noting that copying AI output verbatim into an article or essay is already unethical.

The debate extends beyond fear of being caught. Some critics argue it’s hypocritical for Anthropic to watermark outputs when frontier models are trained on vast amounts of unlicensed human-created content. “Having an AI that watermarks your work is terrifyingly ironic given how many frontier models came by their training data,” one user posted. Still, many Redditors side with the company. As one commenter put it, “There is literally no good argument for why this isn’t a good idea. The only reason you wouldn’t want this is to lie to people.” Anthropic hasn’t disclosed the technical specifics of the watermarking scheme, but it’s expected to use a cryptographic pattern that survives light edits. The policy aligns with broader industry trends around provenance tracking, and similar systems are being tested by OpenAI and Google.