Anthropic to Add Invisible Watermarks to Claude's Text Outputs
HarperSCarroll · x · 2026-08-11
Anthropic is introducing watermarks for its Claude model's text outputs. Similar to Google's SynthID, this mechanism works by slightly shifting the probability distribution of the generated text, making it completely invisible to human readers.
Currently, the detection mechanism for this text watermark exists only within Anthropic, though the company states it is working to enable detection by users and third parties. For image outputs, Claude will add a signature to the metadata, which is notably easy to remove.
More from Models
- Google's Gemini App Hits 1 Billion Monthly Users, Gemma Downloads Reach 1B — OfficialLoganK · 2026-08-12
- Frontier Models Show Huge Gaps in Implicit Financial Knowledge — rickasaurus · 2026-08-12
- Insider Predicts New Model Releases from OpenAI and Anthropic Within Two Months — willdepue · 2026-08-12
- Claude Officially Marks AI Content Steganographically, False Positives Reported — johnnyApplePRNG · 2026-08-12
- Grok's Hallucination: Invents a Hidden 'Elon Only' Settings Menu — altryne · 2026-08-12
- llama.cpp PR Adds Ling-3.0 Support: Devs Report Solid Voice Assistant Results — Public_Umpire_1099 · 2026-08-12