Anthropic embeds invisible watermarks in Claude text, new tool detects them

Anthropic has begun embedding imperceptible watermarks into text generated by its Claude models, a move prompted by the EU AI Act's Code of Practice on Transparency of AI-Generated Content. These watermarks are designed to survive copy-pasting and are applied across all Claude interfaces and integrations, with new models shipping with the mark from August 2, 2026, and older models being retrofitted. A new tool called 'no-watermark' has been developed to detect and strip these invisible characters, as well as other stylistic tells, though it does not address statistical token-distribution watermarking. AI

IMPACT This development highlights increasing efforts to track AI-generated content, potentially impacting content authenticity and detection tools.

RANK_REASON The cluster describes a new tool for detecting and removing watermarks from AI-generated text, which is a specific software utility.

Read on dev.to — LLM tag →

• Anthropic
• AWS
• Claude
• Coalition for Content Provenance and Authenticity
• EU AI Act
• Google.Cloud
• Kirchenbauer
• Microsoft Foundry
• No Watermark

• product
• policy

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →