LSAS Platform Docs

Capabilities and qualification status.

LSAS Platform integrates organizational alignment, deterministic governance assessment, and bounded AI execution. Implemented controls and their evidence are described below; deployment qualification is assessed for the intended environment and use.

Updated 2026-10-31

evaluation

Organizational alignment and governance evaluation

Versioned alignment with independent approvals, durable structured-case assessments, and scoped evidence exports. Approval and activation are distinct from applied runtime enforcement: approved descriptive alignment still requires a supported executable integration before it can authorize runtime behavior. Structural evidence checks do not establish source authenticity, factual truth, or clinical correctness.

evaluation

Runtime policy evaluation

Authenticated applications invoke registered skills through policy-bound grants and constrained transport. Supported checks precede effects and final release. An opt-in, time-bounded nonclinical review profile binds approval to the exact held artifact and original application.

demonstration

Guided sandbox and workflow design

Explore raw versus governed examples, expected fixtures, simulated overrides, and workflow design. These demonstrations are identified separately from measured control results and production approvals.

evaluation

Read-only discovery assessment

Native Windows x64 and macOS Intel packages assess approved files through local LSAS policy checks and keep reports local. Optional SQL Server collection is implemented and has isolated synthetic collector and evaluator evidence; customer connections still require driver setup, access review and qualification. The separate PostgreSQL lsas-scan command performs bounded inventory and sampled email/SSN pattern checks. PostgreSQL also stores results for optional connected reporting; it is not a source type in the file/SQL Server workspace. Assessment findings describe required controls, not applied source remediation or live AI enforcement.

evaluation

Governed skills and specialist agents

Versioned registered skills and a deterministic coordinator enforce scope, destinations, deadlines, and budgets. Specialist proposals cannot expand authority. Authoritative-source retrieval, source-driven policy evolution, and the assessment model broker remain separate unfinished capabilities.

qualification required

Production healthcare and medtech qualification

Customer identity, network and secret controls, recovery, capacity, intended use and independent security/domain assessment require deployment-specific evidence. LSAS Platform does not currently assert certification, clinical validation, or autonomous patient-impact readiness.

Understanding the evidence

A demonstration shows a scenario. A measured test establishes a particular behavior under stated conditions. Production qualification additionally requires evidence for the intended deployment, operating procedures and use. These evidence classes are tracked separately.

This is the public capability summary. Detailed engineering findings and chronological audit records are available to authorized platform administrators.

Runtime architecture

An open invitation

Help shape the future of AI governance.

The challenges organizations face with AI aren’t theoretical. We’re learning from the people navigating them every day.

If you work in governance, risk, compliance, enterprise architecture, security, or AI engineering, we’d welcome your perspective.

Share Your Perspective

A conversation, not a sales pitch.