What the editor can—and cannot—see.
FreeRedact minimizes trust by keeping source documents in an isolated browser context and rebuilding sanitized outputs locally.
Network boundary
The editor loads application code, OCR assets, and face-detection assets from its own origin. It does not load advertising, session replay, third-party analytics, or cloud AI. Document bytes, OCR text, findings, coordinates, filenames, and thumbnails are never sent to the parent shell.
Export boundary
Images are decoded into a fresh canvas, redactions are painted into the pixel buffer, and a new image is encoded without inherited metadata. PDFs are rendered page by page and rebuilt from sanitized page images.
Limitations
Automated detection can miss information. FreeRedact is an assistive review tool, not a legal compliance certification. Password-protected PDFs, video, audio, and files over the documented limits are not supported in the MVP.
Vulnerability reporting
Please report suspected security issues privately to security@freeredact.com. Do not include a real sensitive document in the report.