Claude is watermarking every response — here's what that means if you use AI for writing
Claude now secretly watermarks every AI-generated response. Here's how AI watermarking works, whether it can be removed and what it means for power users.
Anthropic has introduced a new feature for Claude, its AI model, which embeds an invisible signal in all text responses. This move aligns with the European Union's AI Act, aiming to enable the identification of AI-generated content. However, the watermark is not a permanent proof of AI authorship; instead, it subtly alters the statistical patterns in the text, making it detectable by specialized software.
The watermark remains undetectable to human readers and does not alter the document's appearance when copied to other platforms. While the watermark can slightly degrade the quality of the AI-generated text, modern watermarking techniques are sophisticated enough that most users likely wouldn't notice the difference in everyday use.
The watermark remains even if the text is revised by another AI model, but it becomes fragile if the AI-generated response is edited by yet another AI. Anthropic's watermark is a step toward model-level watermarking for AI writing, but it should be considered alongside other evidence rather than as definitive proof of AI authorship.
Written by urgent.news from Tom's Guide's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
