18 August 2026updated 17 August

Anthropic adds invisible watermarks to Claude text for EU compliance

• Anthropic will embed undetectable patterns into Claude's text output to comply with an EU regulation requiring AI companies to mark synthetic content by December.
• The watermark works by slightly constraining random word choices at non-critical moments, like choosing 'grey' or 'overcast' to describe weather, leaving a pattern detectable only with a special key.
• Anthropic says watermarks will not degrade writing quality or increase costs, though some critics argue constraining word choice will force less precise language overall.
• Light editing may preserve the watermark, but completely rewriting text will remove it. Other AI companies including Google's Gemini already use similar watermarking systems.

How it was covered

AI BreakfastIndependent editors

Anthropic released guidance on reducing Claude Code token costs through prompt caching, which cuts repeat input costs to 10 percent, while also testing a side-by-side model comparison interface. The newsletter emphasises that a safety issue exposed 133 million contractor requests due to accidentally disabled biological and chemical weapon classifiers, and notes controversy over Anthropic's new global text watermarking system which critics argue degrades writing quality despite company denials.

Reported by The Guardian, The Verge, TechCrunch