Anthropic to Add Invisible Text Watermarks to New Claude Model Outputs
rohanpaul_ai · x · 2026-08-12
Anthropic announced it will add invisible watermarks to text generated by new Claude models to comply with the EU AI Act's transparency requirements.
- Watermarking Mechanism: Marking occurs at the model level, applying across all Claude products like the API, Claude Code, and major cloud platforms. Text watermarks are embedded directly into the content, surviving copying and some editing.
- Image Metadata: Generated SVG, PNG, and JPG files will include signed C2PA provenance metadata.
- Timeline & Detection: Models launched in the EU on or after August 2, 2026, will support this at launch, with existing models currently in progress. Anthropic is developing detection tools and will release technical docs.
- Limitations: A detected mark cannot prove absolute original authorship by Claude (due to potential edits/summaries), and missing marks prove little, as heavy rewriting or metadata stripping can erase the signal.
Related event: Anthropic to Add Invisible Watermarks to Claude Outputs for EU Compliance(11 posts)→
More from Models
- Anthropic's Computer Use tested: works in Chrome, lacks full desktop control — cedric_chee · 2026-08-12
- Rumor: SpaceXAI to Finalize $60B Cursor Acquisition and Launch Grok 4.6 on August 14 — koltregaskes · 2026-08-12
- Meta Drops 30B Local Open-Source Model as OpenAI Unveils Unrestricted Cybersecurity AI — Dapper-Tale-4021 · 2026-08-12
- Elon Musk: Grok 4.6 to Launch Later This Week After Early Bug Fixes — XFreeze · 2026-08-12
- Mistral Unveils European Compute Units and Regional Inference, Adds Third-Party Model Support — sophiamyang · 2026-08-12
- Gemini's Share of AI-Generated Text Collapses to 1.9% — morqon · 2026-08-12