🔐 Security & Governance

AI Pipeline is designed as a security-first control center for AI agents — with multi-tenant isolation, centralized policy, and deep observability.

Summary: AI Pipeline provides hub / tenant separation, per-tenant runtimes and data planes, role-based access control, and detailed logging. This makes it easier for security teams to understand, approve, and monitor AI usage across the enterprise.

Last updated: 2024-12-02 UTC

🗝️ Architecture & data isolation

Hub / Tenant model

The Hub acts as the control plane for configuration, policies, catalogs, and metrics. Tenants act as isolated execution environments and data planes. This separation is core to how AI Pipeline is built.

• Hub: configuration, policies, auditing, observability
• Tenant: agents, tools, and data access
• Clear boundaries between internal business units or external clients

Per-tenant runtime & storage

Each tenant can have its own runtime, database schema, and object storage namespace. This allows for:

• Strong data isolation
• Tenant-specific performance and scaling characteristics
• Optional region / residency-aware deployment topologies

🧑‍💻 Access control & identity

SSO & identity providers

• Support for enterprise SSO / IdPs (e.g., Okta, Azure AD, Auth0)
• Single sign-on into the AI Pipeline control plane and tenant UIs
• Consistent mapping from IdP groups to platform roles

Role-based access control (RBAC)

• Hub admin, tenant admin, builder, and viewer roles
• Fine-grained permissions for creating, editing, and publishing agents
• Optional separation of duties for higher-risk workflows

📚 Models, tools, and data governance

Model governance

• Configure allowed model providers per environment / tenant
• Enforce model selection policies for specific agents or classes of data
• Support for routing strategies across providers while preserving guardrails

Tool & data governance

• Central catalog of tools (integrations, APIs, internal services)
• Explicit declaration of which tools an agent may call
• Patterns for PII / PHI handling and least-privilege tool access

📊 Logging, audit, & observability

End-to-end logging

• Structured logs for agent invocations, tool calls, and errors
• Comprehensive audit trail for compliance requirements

Metrics & dashboards

• Prometheus metrics for latency, throughput, and error rates
• Grafana dashboards per tenant, agent, and environment
• Support for alerts on anomalous behavior or degraded performance

📜 Compliance-ready posture

AI Pipeline is designed to fit into your compliance program rather than replace it.

• Supports data classification and residency-aware deployment choices
• Facilitates DPIAs / risk assessments for AI usage
• Built to work in regulated industries (finance, healthcare, public sector)

Formal certifications (e.g., SOC 2, ISO 27001, HIPAA) can be pursued depending on deployment model (managed SaaS vs. private deployment).

📄 Need a security brief for review?

Share AI Pipeline with your security team

We maintain a concise 2-page security and architecture overview suitable for vendor security reviews. Contact us at contact@aipipeline.io to request the latest version or to schedule a deep-dive with a security engineer.

📥 Request Security Brief by Email

Please email contact@aipipeline.io with your security questionnaire