VaultSort Launches Guardian: On-Device AI for Finding and Protecting Sensitive Files
Key Takeaways
- ▸Guardian scans for sensitive files (identity documents, financial records, credentials, medical files) entirely on-device with zero network calls or uploads
- ▸On-device OCR and machine learning eliminate the privacy paradox of uploading sensitive data to a cloud service to discover it's exposed
- ▸Analysis reports exist only in memory and never persist to disk, making data breaches technically impossible from the software's architecture
Summary
VaultSort, a Mac file organization and security tool, has launched Guardian, an on-device AI feature that scans for sensitive files—including identity documents, financial records, credentials, and medical files—entirely on the user's Mac with zero network calls or uploads. The feature addresses a fundamental privacy paradox where cloud-based file analysis tools require uploading the sensitive files they're meant to protect.
Guardian uses on-device machine learning and optical character recognition to identify files by content, metadata, and pattern matching. Users can point Guardian at any folders they choose, review results ranked by severity, and integrate findings with VaultSort's encryption tools. Critically, Guardian's analysis report lives only in memory and never persists to disk—the company literally cannot access what Guardian finds because data never leaves the user's machine.
VaultSort 4.4.0 also introduced expanded OCR for image categorization, improved photo grouping, and a redesigned rule builder with content-based scheduling. The company maintains a one-time purchase model without subscriptions; Guardian is free for existing users and costs $24.99 for new customers.
- Integrates with VaultSort's encryption features for end-to-end user control and file protection
- Available as a $24.99 one-time purchase for new users, free for existing VaultSort customers; no subscription model
Editorial Opinion
VaultSort's Guardian solves a critical paradox that has plagued competitors: the absurdity of uploading sensitive files to the cloud to find out they're exposed. By keeping all analysis on-device and never persisting findings to disk, the architecture eliminates both the technical and legal surface for data breaches. The one-time purchase model reinforces privacy-first values over surveillance-based monetization. This could establish a new standard for how consumer security software should handle sensitive data.
