Claude Adds Watermarks To AI Outputs
Claude now embeds imperceptible statistical watermarks in generated text and attaches signed C2PA provenance metadata to supported files. The model-level marking applies across Claude’s apps, API, Claude Code, and cloud distribution channels worldwide.
This is compliance infrastructure disguised as a user-invisible feature, but it raises important questions for developers whose workflows mix human and AI-authored content.
- –Statistical token watermarking can survive copy-paste and light editing, while remaining invisible to readers
- –C2PA metadata gives supported PNG, JPG, and SVG files a tamper-evident provenance trail
- –A detected mark indicates Claude processing, not definitive proof of AI authorship or exclusive machine generation
- –Human drafts sent to Claude for rewriting, translation, or proofreading may inherit an AI mark
- –Detection tools and broader support for older models are still forthcoming
DISCOVERED
1h ago
2026-08-15
PUBLISHED
1h ago
2026-08-15
RELEVANCE
AUTHOR
Better Stack