Anthropic launches imperceptible watermarking to trace Claude-generated content
goyalshaliniuk · x · 2026-08-18
Anthropic has introduced an imperceptible watermark for text generated by supported Claude models. Embedded through subtle statistical patterns in token selection, the signal persists after copying, pasting, or light editing. The rollout covers Claude.ai, the API, Claude Code, and cloud deployments. Additionally, C2PA provenance metadata will be added to supported image and file outputs. Note that the watermark does not prove the entire piece was written by Claude; rewritten or summarized content may also carry the signal.
More from Safety
- Poll: Water and Energy Use Top Reasons for Opposition to Data Centers — soumitrashukla9 · 2026-08-18
- Shanghai AI Lab Paper: Agentic AI Poses Escalating Risks to Human Agency on Three Cognitive Levels — Shanghai-AI-Laboratory · 2026-08-18
- Anthropic-Affiliated Paper Shows "Mind Viruses" Can Spread Between LLM Agents; a System-Prompt Warning Blocks Them — Scobleizer · 2026-08-18
- Wedding speech full of Claudeslop sparks calls for real-time Pangram AirPods — dioscuri · 2026-08-18
- Text Watermark Detection Does Not Require Rerunning the LLM — rasbt · 2026-08-18
- Blog post: Aligned agents can still lead to systemic failure — logangraham · 2026-08-18