Anthropic explains how Claude’s new watermarking works — and what it means for users

Anthropic has published a detailed blog post clarifying how it will watermark text generated by its AI chatbot Claude, a move that has sparked debate among users since the company first announced it earlier this week. The watermarking is part of Anthropic’s compliance with the EU AI Act’s Transparency Code, which requires AI companies to implement systems that make AI-generated content identifiable.

How Claude’s watermarking works

Anthropic explained that when Claude makes “low-stakes choices” — such as selecting between words like “overcast” and “grey” to describe weather — it can embed a pattern in its responses that is “undetectable to the reader, but is detectable to anyone who has a key that encodes it.” The company emphasized that watermarking does not affect output quality: “To a reader, a watermarked response is indistinguishable from an unwatermarked one.”

Anthropic will use the SynthID-Text approach developed by Google DeepMind in 2024 and plans to release a watermark detection API. This method differs from AI detection tools like Pangram, which look for stylistic “tells” in writing. Anthropic noted, “Picking up on these patterns is fundamentally different from checking for a watermark.”

Can the watermark be removed?

Anthropic acknowledged that while light editing probably won’t remove the watermark completely, “a complete rewrite where every word is replaced will.” However, the company added, “In the latter case, of course, it’s arguable whether the text can any longer be described as AI-generated.”

For text that Claude only proofreads or lightly edits, the watermark’s presence depends on “the length of the text and how heavily Claude has edited it.” If only lightly edited, “nearly all the words” come from the human author, leaving “very little (if anything) for the watermark to attach to.”

Impact on code generation

Anthropic said that code will carry less watermarking than other text because the model must produce functional code, limiting its freedom to choose between equally valid options. Still, “in areas where there is an arbitrary choice between particular words or terms within the code, the watermark can be used,” but it will have a “negligible effect” on the actual code produced.

Why this matters to users

The announcement has triggered backlash, with some users on Reddit calling it a “conspiracy against innocent Claude users,” while others argue that opposition suggests a desire to deceive. Business Insider reported that “dozens” of users on X have claimed to cancel their subscriptions. Anthropic also noted that other major AI developers have signed the same Code of Practice and will implement their own watermarks, indicating industry-wide adoption.

Conclusion

Anthropic’s watermarking aims to meet regulatory requirements while minimizing disruption to users. The company’s detailed explanation addresses common concerns about detection, editing, and code generation, but the controversy highlights ongoing tensions between AI transparency and user autonomy. As other developers follow suit, watermarking is likely to become standard practice across the industry.

FAQs

Q1: Will watermarking degrade Claude’s response quality?
Anthropic states that watermarking does not affect output quality; watermarked responses are indistinguishable from unwatermarked ones to readers.

Q2: Can editing remove the watermark?
Light editing probably won’t remove it completely, but a full rewrite that replaces every word will. If only lightly edited, there may be little watermark to detect.

Q3: How does watermarking affect code generated by Claude?
Code will have less watermarking because the model must produce functional code. Where arbitrary choices exist, watermarking may be applied, but it has negligible effect on the code itself.