Product milestones, and the developments in AI provenance that shape what we build.
Industry
Anthropic launches a free Content Credentials checker for Claude files
Anthropic has launched a free browser tool that reads the C2PA Content Credential Claude embeds in image, video, and audio files it creates or edits, reporting only whether the credential is present, not the file’s substance. It covers one of the same signals Provenote reads, but not text — Claude marks text with a separate watermark, and the detection API for that watermark remains in private preview with no public access. Anthropic’s tool.
Industry
Anthropic opens Claude watermark detection in private preview
Anthropic began offering a detection API for its Claude text watermark on 1 September 2026, in private preview rather than general availability. Access is limited to categories it lists as regulators, law enforcement, media, fact-checkers, researchers, educational organisations, EU civil society groups, and enterprises with EU AI Act compliance duties. This is the authoritative provenance signal Provenote has been built to read; we have not been granted access and do not detect Claude watermarks today. Anthropic’s announcement.
Company
Provenote enters private beta
Provenote, a neutral, provenance-first verification layer over AI watermarks and detectors, is now open to a first set of design partners. Request early access.
Company
The Provenote blog is live
Notes on proving how content was made, at least two posts a month, each fact-checked against primary sources before it ships. Read the blog.
Company
Our cardinal rule, in writing
Provenote will not claim anything that cannot be verified, across the product, the site, and the blog. Our principles.
Industry
Anthropic begins watermarking Claude’s text output
Anthropic announced a SynthID-style text watermark for Claude, with a detection API to follow, another verifiable provenance signal Provenote is built to read. Announcement.