Claude Rolls Out Invisible Text Watermarks and C2PA File Provenance
Patient-Pollution46 · reddit · 2026-08-11
Anthropic has implemented invisible watermarking for all Claude-generated text and C2PA signed metadata for files. The text watermark survives copy-pasting and light edits across API, web, and code interfaces, while image files receive provenance signatures to detect tampering. Driven by EU AI Act transparency rules, the move sparks debate on whether AI content will become second-class by default.
More from Models
- Using System One models in Swift: fast, deterministic decisions via Apple Foundation Models — rxwei · 2026-10-03
- Linux Kernel CVEs Surge From ~500 to 1500+ Per Release, LLMs Blamed for Bulk of the Rise — burny_tech · 2026-10-03
- Developer Complains OpenAI's Coding Model Endlessly Scopes Creeps Instead of Finishing Tasks — DavidWells · 2026-10-03
- Sonnet 5 Spotted in Google Antigravity Backend, Which Still Runs Sonnet 4.6 — brandon_galang · 2026-10-03
- Every's Dev Day chat with Matthew Berman: budget gone the moment he tried Ultrafast — every · 2026-10-03
- Wish list: a Qwen4 27B with 100B+ Engram offloaded to RAM and NVMe for local users — casper_hansen_ · 2026-10-03