Resource access. Identity. Compliance.
The independent Resource Access Gate handles resource access. AI-IAM returns ALLOW / DENY / STEP_UP within a scoped enterprise deployment. The Compliance Engine returns PASS / WARN / FLAG / BLOCK.
LogionOS Platform
The runtime control plane for compliant AI.
Evaluate resource access, supported identity context, and query or content compliance through explicit, ordered runtime stages.
Enforcement lifecycle
Resolve each boundary in sequence.
Capture supported context
Accept query or content, jurisdiction, agent identity, PII or data classification, and supported agent_role / resource_tags.
Apply Resource Access Gate
Independently evaluate resource_tags with agent_role at the resource-access boundary, before any identity or content evaluation.
Apply AI-IAM
Check agent identity, status, jurisdiction, clearance, and rate budget; return ALLOW, DENY, or STEP_UP. Available only within a scoped enterprise deployment.
Evaluate compliance
Classify query or content, match shipped rules, and return PASS, WARN, FLAG, or BLOCK with the judge result.
Record delivered evidence
Write request_id, query hash, optional IAM effect, classification, results, prev hash, and chain hash.
Four-stage runtime architecture
Distinct controls, connected in sequence.
Stage 1 - Resource Access Gate
Evaluate resource_tags with agent_role as an independent resource-access boundary. This stage runs first.
Stage 2 - AI-IAM Enterprise deployment scope
Check agent identity, status, jurisdiction, clearance, and rate budget; return ALLOW, DENY, or STEP_UP. This is a reference implementation available within a scoped enterprise deployment, not the default path of the current public API.
Stage 3 - Compliance Engine
Evaluate query or content and policy matches; return PASS, WARN, FLAG, or BLOCK with classification and judge result.
Stage 4 - Evidence ledger
Link each record with prev hash and chain hash for cryptographic continuity after the compliance decision resolves.
Policy packs
Start from mapped controls. Add your operating rules.
Activate pre-built packs by regulation, jurisdiction, and industry, then layer organization-specific requirements without changing application code.
Privacy and data
HIPAA-aware, GDPR, APPI, LGPD, PIPA, PIPEDA, and cross-border data handling profiles.
AI governance
EU AI Act, NIST AI RMF, model-use restrictions, human oversight, disclosure, and risk-class controls.
Industry operations
Healthcare, financial services, legal, telecommunications, and enterprise data-handling profiles.
Deployment options
Enforce policy where your AI runs.
LogionOS Cloud
Managed control plane and runtime endpoints for the fastest path to production.
Private cloud or VPC
Keep processing within your cloud account and approved regional boundary.
On-premises
Run the enforcement plane inside your private network alongside regulated workloads.
Air-gapped
Operate without external network dependency for restricted and high-assurance environments.
Evidence model
Delivered fields, linked record by record.
Each stored record is tamper-evident and hash-chained through the delivered evidence fields.
Request
request_id and query hash identify the stored request without claiming additional context fields.
Runtime results
IAM effect when agent_id is provided, classification, compliance result, and judge result.
Chain continuity
prev hash links the prior record; chain hash records the current chain value.
API and SDK integration
Add the runtime gate without rebuilding your stack.
Compliance API
Submit supported query or content and consume the separate runtime results.
Python SDK
Submit supported query or content and runtime context from Python services.
JavaScript / TypeScript SDK
Submit supported query or content and runtime context from JavaScript services.
Proxy and middleware
Connect existing model traffic through the documented integration path.
Deploy your first control
Turn AI policy into runtime behavior.
Choose one production workflow. We will help map the supported inputs, integrate enforcement, and validate the delivered evidence fields.