Anthropic Details Text Watermarking: No Quality Impact, No Hidden Characters
arpit_bhayani · x · 2026-08-15
Anthropic published a blog post detailing how Claude's text watermarking works and the compliance background behind it.
- Technical Mechanism: LLMs select tokens from a probability distribution. The watermarking technique guides the model to make specific choices at points where the meaning remains consistent to the reader (e.g., choosing between synonyms), embedding a statistical signal. It requires no extra tokens, incurs no additional cost, and is undetectable to readers.
- Privacy & Tracking: The watermark carries no identifying information and cannot be traced to specific individuals, organizations, or chat sessions.
- Policy Driven: The move is to comply with the EU AI Act. As of August 2, the EU requires AI providers to mark AI-generated content; other major developers are also signing the Code of Practice to implement their own watermarking solutions.
Related event: Anthropic Deploys Text Watermarking for Claude to Comply with EU AI Act(31 posts)→
More from Safety
- Linux Kernel CVEs Surge From ~500 to 1500+ Per Release, LLMs Blamed for Bulk of the Rise — burny_tech · 2026-10-03
- COLM 2026 Launches DAIH Workshop on Deploying LLMs/VLMs Responsibly in Healthcare — StellaLisy · 2026-10-03
- Trillium Labs wants to do open research on recursive self-improvement and agents — nordicinst · 2026-10-03
- Trillium Labs Wants to Research Self-Improvement and Model Behavior in the Open — Wired AI · 2026-10-03
- Cloudflare Turnstile everywhere: anti-AI scraping walls now hit human users — sethlazar · 2026-10-02
- Filler tokens let frontier models reason invisibly: 13-point gains undetectable by CoT monitoring — PandaAshwinee · 2026-10-02