Private legal AI
Private legal AI is an architecture, not a privacy adjective
Private legal AI for law firms: compare shared cloud, single-tenant and private-perimeter deployment for sensitive legal work.
Quick answer
For a law firm, 'private legal AI' should describe the actual deployment boundary: tenancy, identity, data stores, model routes, network egress, retention, audit and administration. A strong private deployment is one the firm's security team can diagram and verify, not simply a product labelled private.
Reviewed September 2026
Deployment models
Three common architectures
| Model | Typical characteristics | What to verify |
|---|---|---|
| Shared enterprise cloud | Vendor-operated cloud application with logical tenant separation, regional hosting and enterprise controls. | Subprocessors, model endpoints, retention, data location, logical isolation and contractual commitments. |
| Single-tenant cloud | Dedicated application/data plane for one customer, often within a vendor-managed or customer-aligned cloud boundary. | Whether inference still leaves the tenant, who administers the environment, key control and network paths. |
| Private perimeter | Application, matter data and approved inference run inside a defined private boundary with outbound paths tightly restricted or denied. | Whether the model actually runs inside the boundary, whether any telemetry exits, upgrade path, support access and attestation. |
Sanctum Lex
Why deployment is part of the product boundary
Sanctum Lex is designed so enterprise identity, matter permissions, data handling, model routing and audit can be governed together. The same lawyer experience can be deployed with different approved inference routes depending on the firm's policy.
Where a deployment uses an approved external model endpoint, Sanctum describes that route as governed egress. Where outbound model routes are denied and inference runs inside the private perimeter, Sanctum can describe the configured boundary as zero egress.
Frequently asked questions
What is private legal AI?
Private legal AI is a legal AI deployment in which the firm's data, identity, model routes and system controls are constrained to a defined environment rather than treated as ordinary multi-tenant application traffic. The exact architecture can range from single-tenant cloud to a private perimeter with local inference.
Is single-tenant the same as zero egress?
No. A single-tenant application can still call an external model endpoint. Zero egress is a network property: no outbound model traffic leaves the defined boundary. Sanctum Lex uses the term only for deployments where outbound routes are denied and inference runs locally or inside the sealed private environment.
Why would a law firm need private legal AI?
Some firms or matters have stricter confidentiality, client, regulatory, data-location or infrastructure requirements than ordinary SaaS controls can satisfy. A private architecture can reduce the number of external trust boundaries and make the inference path easier to explain to clients and security teams.
Continue comparing
Evaluate on your own matter