Trust6 is a desktop application (macOS and Windows) built for consultants and professionals who work with confidential client data: legal files, financial records, HR matters, advisory mandates.

It lets you use large AI models (Claude, GPT and others) on your files, without ever exposing your clients’ information to the providers of those models.

How that is possible

Everything rests on local pseudonymization. Before a message leaves your machine:

  1. Trust6 detects sensitive information (names, organizations, emails, phone numbers, amounts, places) with an analysis engine that runs entirely on your computer.
  2. Each value is replaced by a neutral, stable token such as PERSON_001 or ORG_002.
  3. You review and approve the result, then only the masked text is sent to the AI model.
  4. When the answer comes back, Trust6 restores the real values on screen, for you only. The mapping table between tokens and real values never leaves your machine.

The AI provider never sees “Jean Tremblay offers $25,000.50”, only “PERSON_001 offers AMOUNT_001”.

One vault per client

Each client gets its own vault: an encrypted database holding that client’s conversations, files and mapping table. Vaults are isolated from one another by encryption: a token created in one client’s vault literally does not exist in another’s. The encryption key lives in your system’s secure keychain, never on disk.

What Trust6 never does

  • Send a native file to an AI model: documents are read and converted to text locally, then masked like everything else.
  • Send anything without analysis: if detection fails, sending is blocked. There is never a silent fallback to clear text.
  • Keep your content on a server: the account server manages your licence, not your data.

Where to go next