I Built an Open-Source Toolkit for AI Watermark & Provenance Forensics
Description and use
ClaudeMark is a local, open-source toolkit for AI watermark and provenance forensics that aggregates statistical signals, invisible Unicode analysis, metadata inspection, security auditing, and sanitization, enabling developers to analyze AI-generated content locally.
Problem
Investigating AI-generated content for provenance, metadata, and hidden signals without sending files to external services.
Outcome
ClaudeMark enables developers to locally inspect AI-generated artifacts, identify forensic signals, audit security, and sanitize metadata while keeping data on the machine.
Workflow
- 1. Install ClaudeMark from GitHub.
- 2. Run the CLI or API to input files or content.
- 3. The detection pipeline analyzes AI signals, Unicode characters, and metadata/provenance layers.
- 4. Review forensic results and optional security audit findings.
- 5. Optionally sanitize selected metadata artifacts before further use.
Evidence
ClaudeMark is an open-source, local-first toolkit for AI watermark and provenance forensics.
The goal is simple: Your files stay on your machine.
- Record date: 2026-09-08T17:01:02.619Z