Liminal AI —
Secure Generative AI Enablement & Governance

Liminal AI is a secure AI enablement and generative-AI governance platform delivering single-pane access to leading models (GPT, Claude, Gemini and others), policy-driven sensitive-data protection on prompts and responses, custom multi-model assistants, in-application AI access across the productivity stack, an SDK for embedding the same controls into custom applications, and full observability of AI usage — fitting operating models whose workforce is already using generative AI and whose security, privacy and compliance posture cannot continue absorbing unmanaged Shadow AI exposure.

A policy engine inspects every prompt and response in real time and applies redact, mask, warn or pass-through actions on entities such as PII, PHI, financial identifiers and internal terminology — with the option to rehydrate masked terms after the model responds — fitting operating models whose regulated data (HIPAA, GLBA, FERPA, PCI-adjacent, IP) cannot be sent verbatim into third-party LLM endpoints, and whose privacy posture demands a deterministic control plane rather than user-side discipline.

Liminal provides a sanctioned, governed surface for AI usage across web, desktop and embedded applications — combined with observability over which users, models and applications are in use — fitting operating models whose security posture has discovered employees pasting sensitive content into consumer AI tools, and whose governance posture cannot rely on blocking AI as a sustainable strategy.

Each customer is deployed in a single-tenant VPC with a dedicated relational database and data lake inside the private subnet; data is encrypted at rest with AES-256 rotating keys and in transit at TLS 1.2+; SSO is delivered via SAML against major IdPs and SDK access via OAuth 2.0 client credentials — fitting operating models whose enterprise security posture demands tenant isolation, IdP integration and auditable cryptography rather than shared-tenant SaaS defaults.

A web workspace (Spaces) covers direct chat with multiple models; a desktop helper extends the same controls into Microsoft 365, Google Workspace, Salesforce, Slack, Adobe and other applications; an SDK (Python, TypeScript, .NET) embeds the same protections into applications you build — fitting operating models whose AI strategy spans both buy-side productivity and build-side application development, and whose architecture posture cannot tolerate a different control plane for each surface.

Your contract is with Liminal AI. Fibi remains your independent advisor — escalating support cases, revisiting AI-governance scope at renewal, and re-comparing Liminal against alternative secure-AI and AI-governance providers at contract events. There is no advisory fee — Fibi is provider-funded.