Software capability
The application supports a single-tenant deployment, local document processing, client-selected identity and a client-approved model route.
Security / Architecture whitepaper
Deployment topology, isolation boundaries, key handling, and the control mapping. Watermarked, shared under NDA, and written for reviewers rather than buyers.
Built for enterprise deployment
The public site describes capabilities. The signed deployment record describes which controls are actually operating for a client.
The application supports a single-tenant deployment, local document processing, client-selected identity and a client-approved model route.
A control is described as active only when it is configured and tested in the named client environment. The public Vercel demonstration is not an air-gapped enclave.
Zero egress applies only where outbound network paths are denied and inference runs locally. An external model API is governed egress, not zero egress.
An integration boundary is not Connected until provider authorization and a successful client-environment test have been recorded.
Enterprise review
The answers are not slogans. Each one maps to a software capability, a deployment condition, or evidence your technical team can inspect.
01
Sanctum Lex is designed for environments where confidentiality and control are fundamental. The platform is built around isolation, least-privilege access, auditable activity and controlled data flows. Our infrastructure team includes engineers with more than 30 years of Linux and systems experience.
02
Firm documents remain within the deployment boundaries agreed with the client. Knowledge access is permission-aware, and Sanctum Lex can support highly restricted or zero-egress operation where outbound paths are actually denied and inference is local.
03
We do not present AI output as infallible. Sanctum Lex can ground analysis in approved sources, expose the underlying material for review, and use adversarial roles to challenge a conclusion rather than simply reinforce the first answer.
04
Access is governed at user and organization level. Firms can determine which people may access particular matters, knowledge sets and capabilities, with enterprise identity and role boundaries available for configured deployments.
05
Yes. Administrators can govern users, permissions, integrations, knowledge sources, model routes, retention and platform policy. The public site distinguishes software capability from the controls actually configured and verified for a client.
06
The lawyer remains responsible for professional judgment. Opposing Counsel and Judicial Review can deliberately challenge a position, surface unresolved issues and expose assumptions before the work leaves the firm.
Contents
Seven sections, roughly forty pages, with diagrams your infrastructure team can build against. We do not publish it because the deployment detail of a client enclave is not public information.
Node specification, network placement, and how the enclave sits inside your existing segmentation.
Process, namespace, and storage separation between the runtime, the index, and the corpus.
Where keys are generated, where they are held, and what happens to them at teardown.
How classification is applied at ingest and enforced before inference reaches a passage.
Namespace restrictions inherited by background jobs, and why an agent cannot exceed the session.
Each guarantee mapped to the GDPR, HIPAA, and SOC 2 control it addresses.
Model updates in place, RAM wipe on teardown, and the evidence produced by each.
The isolation model, the threat cases we designed against, and what we deliberately did not build.
Hardware requirements, provisioning, and where the enclave touches your network.
The control mapping and the data handling statement, ready to attach to a security questionnaire.
Tell us who is reviewing and we will send the whitepaper under NDA, then walk it with them.