Trust center
The questions your risk committee will ask, answered up front.
This page is maintained by Vanguard Command OS and describes controls that are live in the product today. It is a description of practice, not an independent certification.
Trust through visibility
Understand the controls behind the interface.
Review access, attribution, audit evidence, and the responsibilities shared with your firm. The artwork represents that structure; it is not a certification seal.
Authentication & roles
Email and password or Google sign-in, with roles stored in a dedicated table and evaluated server-side. Role checks never rely on browser state.
Row-level access control
Every table carrying firm or client data enforces access policies in the database. A request for another firm's row fails at the data layer, not in the UI.
Tamper-evident audit chain
Audit entries are hash-chained so edits or deletions are detectable. Eligible plans seal periodic signed Merkle roots for independent verification.
AI attribution & disclosure
Every AI-assisted artifact carries the agent, model, approver, and timestamp. Firms can export the full provenance record for a matter.
Secret handling
Integration credentials, signing keys, and webhook secrets are held server-side and never exposed to the browser. Inbound webhooks are signature-verified.
Encryption in transit & at rest
All traffic is served over TLS, and stored data and documents are encrypted at rest by the managed platform.
Shared responsibility
Vanguard operates
The platform, access enforcement, audit integrity, model routing, and integration security.
Your firm decides
Oversight thresholds, who holds which role, retention windows, which integrations are enabled, and what clients can see.
Both confirm
That professional responsibility obligations, court rules, and client engagement terms are satisfied for each use of the system.
Outside services we use
Last reviewed: 2026-10-15
| Service | Purpose | Data received | Region | Privacy |
|---|---|---|---|---|
| Lovable Cloud | Application hosting, database, sign-in, file storage | All firm and client records | United States | Policy |
| Cloudflare | Edge hosting and network delivery | Request traffic in transit | Global | Policy |
| OpenAI | AI drafting, analysis, embeddings | Prompts and matter excerpts sent for a task | United States | Policy |
| Google (Gemini, Workspace) | AI failover; attorney Gmail and Calendar when connected | Prompts; email and calendar items the attorney authorizes | Global | Policy |
| Microsoft (Outlook, Teams, OneDrive) | Attorney email, calendar and files when connected | Items the attorney authorizes | Global | Policy |
| Composio | Per-attorney sign-in to workplace tools | Tool tokens and the requested tool actions | United States | Policy |
| SendGrid (Twilio) | Fallback and firm email delivery | Recipient addresses and message content | United States | Policy |
| DocuSign | Electronic signature envelopes | Documents and signer names/emails | United States | Policy |
| Adobe Acrobat Sign | Alternative e-signature | Documents and signer details | United States | Policy |
| Browserbase | Supervised browser sessions for public records and form pre-fill | Pages visited, entered form data, session recordings | United States | Policy |
| LlamaIndex (LlamaParse) | Document parsing | Uploaded documents being parsed | United States | Policy |
| Vaquill | Read-only legal research | Research queries | United States | Policy |
| Legal Data Hunter | Read-only legal data search | Search queries | United States | Policy |
| Paddle | Subscription billing | Billing contact and payment details | United Kingdom / United States | Policy |
Common questions
- Is client data used to train models?
- No. Prompts and matter content are sent to the model provider only to produce the requested output for your firm, and are not contributed to model training by Vanguard.
- Who can see a given matter?
- Firm users with a role granting access to that matter, and the client contacts explicitly attached to it. Access decisions are enforced in the database and recorded in the audit trail.
- How long is data retained?
- Retention is configured per firm to match your records policy. Audit entries are retained for the full contract term so the chain remains verifiable.
- Do you hold formal certifications?
- Vanguard is not making a certification claim on this page. Ask us for our current attestation status and the security questionnaire response pack during procurement.
AI-first, partner-accountable
Send us your questionnaire
We answer security reviews in-house, in writing, and usually within three business days.
- 57 specialist agents
- 3 oversight tiers
- Hash-chained audit trail
- EN / ES parity