Security and privacy for AI workloads
Lucenia is designed from the ground up for organizations that cannot compromise on data privacy. Every component of the AI retrieval pipeline runs on your own infrastructure, and enterprise-grade security controls ensure that sensitive data is protected at every layer.
100% private deployment
Unlike cloud-only AI search services, Lucenia deploys entirely on your infrastructure:
- Self-hosted: Run on bare metal, VMs, Docker, or Kubernetes with Helm charts
- Private model connections: Connect to embedding and inference models running in your own VPC via AWS Bedrock or self-hosted HTTP endpoints — your data never leaves your network
- No external dependencies: The content extraction, chunking, and indexing pipeline requires zero external API calls (embedding is the only step that calls a model, and that can be self-hosted)
Attribute-based access control (ABAC)
Lucenia supports attribute-based access control with policy-driven field redaction, ensuring that AI retrieval results respect access policies:
- Document-level security: Restrict which documents a user or role can access. Search results automatically exclude documents the user isn't authorized to see — even when retrieved through vector search.
- Field-level security: Control which fields are returned in search results. Sensitive fields (PII, classified content, internal metadata) can be hidden from unauthorized users while still allowing search over the remaining fields.
- Field masking: Hash or redact sensitive field values in search results. Users can search for documents but see masked values for fields they don't have clearance to read in plain text.
These controls apply to all search methods — lexical, vector, hybrid, and pipeline-processed results — ensuring consistent security across AI retrieval workflows.
Compliance
Lucenia is built to meet the security and compliance requirements of the most demanding organizations:
| Standard | Description |
|---|---|
| SOC-2 | Lucenia supports SOC-2 compliance requirements with comprehensive audit logging, access controls, and data protection capabilities |
| FIPS 140-2/3 | FIPS-compliant cryptographic modules for TLS encryption, certificate management, and authentication backends |
| Federal | Designed for federal and government workloads with FIPS compliance, ABAC security, air-gapped deployment support, and private model integration |
Lucenia's security architecture — combining FIPS-compliant encryption, attribute-based access control, field-level redaction, and fully private deployment — makes it uniquely suited for regulated industries, government agencies, and defense applications where data sovereignty is non-negotiable.
Private model integration
For organizations that require complete data isolation, Lucenia supports fully private model deployments:
| Provider | Privacy level | Configuration |
|---|---|---|
| AWS Bedrock | Data stays in your VPC | Configure region and IAM credentials |
| Self-hosted HTTP | Fully air-gapped | Point to any model with a REST endpoint |
| OpenAI | Data sent to OpenAI API | For non-sensitive workloads |
The HTTP provider supports any embedding or inference model with a REST API, enabling integration with privately hosted models — including open-source models running on your own GPU clusters.
Role-based model access
The ML Commons model access control feature lets you restrict which users and roles can access specific models. This is critical in multi-tenant environments where different teams may have access to different models with different cost and capability profiles.