Anthropic explains how Claude's text watermark actually works

If you draft with Claude, your text will carry an invisible statistical fingerprint—here's what that means in practice.

This story ran in the Coral Wire edition of Fri, Aug 21 2026, when the Coral Index stood at 57/100 (Mixed signals).

Coral Wire selects stories by a Claude-graded relevance score and writes each headline and one-line summary with an Anthropic Claude model from the linked source. Story gathering is not AI: it is a fixed set of feeds and dockets. The linked original is always the authority.

Also in this edition

• EU rules AI training on your work without consentWire — New regulations limit how AI systems can use author content for training data.
• One-third of web now shows AI-generated content signaturesCommunity — Market saturation threatens author differentiation; human authenticity becomes competitive edge.
• YouTube creators face backlash over AI sponsorship dealsAI — Authors face identical trust and disclosure dilemmas accepting AI partnerships.
• ChatGPT search expands with targeted source capabilityAI — AI research tools evolve; authors must adapt research workflows accordingly.
• Court tosses Hachette's AI book cover copyright claimsAI — Legal precedent clarifies current AI copyright landscape for authors and publishers.
• EU adopts Voss Report protecting author copyrights from AIWire — Major policy victory establishes enforceable copyright standards against unauthorized AI training.
• Texas escalates book bans; authors face renewed censorship pressureCommunity — Direct threat to author readership, market access, and creative freedom expands.
• Fourteen writer federations unite demanding stronger AI protections globallyWire — International advocacy movement gains institutional momentum pushing regulatory protections for author rights worldwide.