Multi-Tenant Policy Packs

Isolated policy, per organization

Each tenant gets its own isolated policy set. Shared baseline rules are read-only; tenant-specific rules can only ADD restrictions — never weaken the platform's immutable controls.

Multi-Tenant Policy Packs

Isolated policy, per organization

Each tenant gets its own isolated policy set. Shared baseline rules are read-only; tenant-specific rules can only ADD restrictions — never weaken the platform's immutable controls.

🔒

Tenant isolation

Policy, decisions, and audit are scoped per org. The auth-derived tenant id is injected server-side — never trusted from the request body.

📦

Versioned packs

Reference packs like lending_v1 ship ready to use; custom packs are versioned so every decision cites the exact policy that fired.

Add-only customization

Tenants tighten, never loosen. Immutable safety rules and red lines hold regardless of tenant configuration.

The Isolation Model

Why add-only matters

In a multi-tenant control plane, the dangerous failure is one tenant's configuration weakening a guarantee that should hold for everyone. The policy model is built so that cannot happen.

01

Shared baseline, read-only

Platform-level safety rules and ethical red lines are immutable. No tenant configuration can remove or override them.

02

Tenant rules add restrictions only

A tenant can make its own policy stricter — more reasons to block — but cannot make it more permissive than the baseline allows.

03

Tenant id is server-derived

The tenant a decision belongs to comes from the authenticated session, never from the request body — so a caller cannot evaluate against another org's policy.

04

Versioned for replay

Every decision cites the exact policy version that produced it, so an examiner can reproduce the verdict under the rules that were actually in force.

FAQ

Common questions

How are tenants isolated in EVE CoreGuard?
Each organization has its own isolated policy set, decisions, and audit trail. The tenant id is derived from the authenticated session server-side, never trusted from the request body, so one caller cannot evaluate against another org's policy.
Can a tenant weaken the platform's safety rules?
No. Shared baseline rules and ethical red lines are read-only and immutable. Tenant-specific rules are add-only — they can tighten policy but never loosen it below the platform baseline.
How does a decision cite which policy applied?
Policy packs are versioned (for example, lending_v1). Every decision records the exact policy version that produced the verdict, so the decision can be reproduced under the rules that were in force at the time.

Request a Design Partner Pilot

Put EVE CoreGuard in front of one real, high-stakes AI workflow. We'll stand up a policy pack, wire the gate, and show you blocked actions with signed evidence.