TrustAI · vault
TrustAI Vault: govern AI actions and mask sensitive data before the model
Classic DLP watches files. TrustAI Vault protects what employees paste into AI — and what agents try to do next — with risk classes and human approval.
Problem
Teams paste contracts, CRM rows and HR notes into ChatGPT. Without a Vault, personal and confidential data can leave the company in one click.
Promise
Detect and mask sensitive entities before any model call. Apply policy tiers for agent tools. Keep evidence for security and legal reviews across Switzerland, the EU and international teams.
Features
Mask before the model
Names, emails, IBANs, health markers and secrets tokenised prior to LLM exposure.
Policy R0–R4
Read, draft, send and high-impact actions mapped to human approval requirements.
Audit-ready logs
Exportable trails for GDPR accountability and Swiss FADP documentation.
Works with Crew
Same Trust Layer for chat and multi-agent missions — not a bolt-on scanner only.
FAQ
Is an AI Vault the same as DLP?
Related, but a Vault is specialised for LLM prompts, agent tool calls and policy classes around AI actions.
Does the Vault block productivity?
No. Structure remains; only sensitive values are masked so models stay useful without clear-text secrets.