Executive demonstration
For investment principals, family offices and investment committees.
A private session on synthetic data: one decision from evidence to outcome, and what protects it today.
Request a Private DemoLIBA Sovereign Intelligence™ · Architectural vision
LIBA SI™ protects your decision records with organization-level isolation and role-based access. We’re building customer-managed keys and customer-defined AI policies to give your institution more control over how its data is accessed and processed.
Four pillars
Each ring is a pillar; each segment, one control in its current status. Select a pillar to see its controls and technical limits.
Decision Integrity & Provenance
Pillar 01
Each organization’s records sit behind their own boundary, under role-based authority.
Each organization is its own tenant. Its records carry its organization key, the database enforces row-level security on tenant tables, and the server checks the organization of every request against the signed-in person’s verified membership.
Demonstration and evaluation workspaces hold synthetic data only and refuse real content.
The complete vault architecture is in development. Existing tenant controls do not mean exclusive customer key custody, a private vector index or protection from an administrator with infrastructure access.
Pillar 02
Encryption in transit today; keys held in your own key service are in development.
The public website and application endpoints use HTTPS. Stored data sits in a managed cloud database and object storage run by the hosting provider, and LIBA, as the service operator, can technically reach it.
Customer-managed encryption is in development. The design uses organization-specific data keys wrapped by a customer-managed key, with authorization, rotation and audit at the key service. A production integration, recovery process and customer approval of that trust boundary are still required. Plaintext remains accessible wherever an authorized application processes it; key ownership alone does not exclude LIBA administrators.
Pillar 03
Engines calculate, AI advises, and a named person decides.
Deterministic engines calculate every figure and AI reviews the analysis as advice. A model can recommend investigating, watching or passing; only a named person can approve. Which providers and models run is set on the server, and the model behind each analysis is recorded with it.
Customer AI policy enforcement is in development. Its tested implementation candidate denies external processing unless the organization permits the exact provider, model and purpose and the provider’s processing conditions have been reviewed. Deployment acceptance is still required. Processing location, retention and training restrictions depend on enforceable provider settings and agreements; a policy label alone cannot establish them.
Pillar 04
Access granted through roles is withdrawn the same way.
Access is granted through roles, so it can be withdrawn the same way. Future application requests require current authorization; an already-issued signed document link may remain valid for up to 60 seconds.
Export and deletion at the end of a subscription follow the terms of your order or agreement, and the available export scope is confirmed for each deployment. Copies held in backups are not deleted immediately when access ends. Revocation blocks future authorized access; it cannot erase plaintext already disclosed or reverse completed processing.
Every assumption, approval, override and outcome becomes part of your institution’s memory. Over time the record shows which assumptions held, which did not, and who decided.
How your data would move in the full design, and which controls protect it today.
Powered by Continuous Intelligence
Private intelligence
Your organization’s documents, decisions, assumptions, outcomes, prompts, corrections, AI outputs, embeddings and usage-derived content are processed only to provide the service to your organization. They are excluded from shared product improvement.
NO AUTOMATIC PRIVATE DATA SHARING.
LIBA Intelligence Evolution™
A future contribution program would need explicit consent covering purpose, data scope and retention, with an audit record and withdrawal of future contributions; any irreversible downstream effects would be disclosed. No such program is available today.
Not offered today: Federated learning, Differential privacy, Customer-specific model training. LIBA SI™ makes no claim to any of them.
The vault above shows each pillar’s controls. Here are the limits that apply to all of them, and what is on the roadmap.
LIBA can access plaintext within the service’s operational trust boundary, and customers do not hold exclusive custody of encryption keys. LIBA SI™ is not a zero-knowledge system and holds no SOC 2 report. Production acceptance for external customer deployments is not yet complete.
On the roadmapCustomer-managed keys · Key rotation under your control · Customer AI processing policy · Attested intelligence execution · Key-based revocationPlanned
Operational control supported by the required verification evidence.
Implemented or undergoing verification, with remaining acceptance requirements.
Approved architectural or engineering direction that is not yet operational.
Under LIBA SI™’s terms, you keep ownership of the data, documents and records your team submits; LIBA SI™ receives a limited right to process them only to provide, secure and support the service.
For investment principals, family offices and investment committees.
A private session on synthetic data: one decision from evidence to outcome, and what protects it today.
Request a Private DemoFor CISOs, CTOs and enterprise procurement teams.
Tenant isolation, roles, sign-in rules, AI configuration and revocation, with the evidence behind each status.
Request a Security Review