Policies Overview

Policies define how AI is controlled in your organization.

They determine which models can be used, what data can be accessed, how integrations behave, and what actions are allowed.

Without policies, visibility does not translate into control.

What policies control

Policies can govern:

Model usage

  • Allowed providers and models

  • Routing logic

  • BYOK requirements

Data access

  • What data can be sent to AI systems

  • Which sources are restricted

  • Context-based filtering

Integrations

  • Which systems can be accessed

  • What actions are permitted

Workflows and actions

  • When approvals are required

  • What actions can execute automatically

Policy-based routing

Peridot supports policy-based routing of AI requests.

This allows you to:

  • Route sensitive workloads to approved models

  • Enforce provider restrictions

  • Separate environments (dev vs production)

Routing decisions are made centrally, not within individual applications.

Enforcement

Policies are enforced in real time.

Enforcement actions may include:

  • Blocking requests

  • Re-routing to approved systems

  • Triggering human approvals

  • Logging violations

Policies as a control layer

Policies operate across:

  • AI Inventory

  • Data Flows

  • Applications

  • Integrations

This ensures that control is consistent across the entire AI environment.

Why this matters

Most organizations define policies but cannot enforce them.

Peridot turns policies into executable control.

What to do next

  • Create your first policy in Creating Policies

  • Connect policies to workflows using Enforcement Actions


Was this article helpful?