Zero-egress legal AI
Zero egress is not the same as zero retention
Zero-egress legal AI explained: how zero egress differs from zero retention, data residency and private cloud for law firms.
Quick answer
A legal AI system has zero egress only when the matter content and inference requests stay inside the defined deployment boundary because outbound model routes are denied. An external API with a zero-retention contract may be appropriate and well governed, but it is still egress.
Reviewed September 2026
Definitions
Four controls that are often confused
| Control | Meaning | Does data leave the boundary? |
|---|---|---|
| Data residency | Data is hosted or processed in a specified region. | Potentially yes. |
| Zero retention | An external processor does not retain request content after processing under the relevant terms/configuration. | Yes, if the processor is external. |
| Governed egress | Outbound inference is allowed only to approved model routes under policy and contractual controls. | Yes, deliberately and audibly. |
| Zero egress | Outbound model paths are denied and inference occurs inside the defined private boundary. | No, for the governed inference path. |
Why it matters
Legal confidentiality questions are architectural questions
A law firm may be comfortable with a well-controlled regional cloud provider for most matters and still have individual clients or engagements that require a stricter boundary. The useful architecture is therefore not ideological: it should let the firm choose an approved route based on matter sensitivity and policy.
Sanctum Lex separates these states so lawyers and security teams are not given a decorative 'private' badge. The environment should be able to state which inference boundary is active and what that means.
Frequently asked questions
What does zero egress mean in legal AI?
Zero egress means matter content and model requests do not leave the defined deployment boundary during normal AI inference because outbound model routes are denied. It is a network and architecture property, not merely a retention policy.
Is zero retention the same as zero egress?
No. Zero retention can mean an external provider does not persist the request after processing, but the request still left the firm's environment. Zero egress means the request did not leave the defined boundary in the first place.
Can Sanctum Lex use external models?
Yes, where the firm's approved policy allows governed external model routes. Sanctum's model gateway is designed to route to approved providers or private endpoints. Those external routes are not described as zero egress.
Continue comparing
Evaluate on your own matter