---
title: Do you need a separate AI governance deployment per region?
url: https://opencontroller.lyzr.com/blog/one-or-many-ai-governance-deployments-across-regions
published: 2026-11-11
updated: 2026-11-11
author: Lyzr
---

# Do you need a separate AI governance deployment per region?

**Usually not. Most organisations can run one governance deployment for inventory, policy and reporting, and place only the parts that touch prompt content (gateways and log stores) in each region. You need fully separate regional deployments when a regulator or contract forbids governance metadata itself from leaving a jurisdiction, or when regions must be operated independently.**

This question comes up in almost every enterprise evaluation, often alongside development, test and production environments. The answer depends less on the tool and more on what data the governance layer holds and where it has to travel.

## What data does an AI governance layer actually hold?

Split it into two planes, because they have very different residency profiles.

| Plane | Examples | Contains personal or customer content? |
|---|---|---|
| Control plane | Agent inventory, owners, versions, policies, posture findings, evaluation scores, cost totals | Rarely. Mostly metadata about agents |
| Data plane | Prompts, responses, retrieved documents, traces with tool arguments | Often. This is where residency rules bite |

If prompts and responses stay in-region, the control plane can often be central. Data protection teams will still want to check the metadata, because owner names and user identifiers are personal data under GDPR.

## What deployment patterns are there?

| Pattern | How it works | Good for | Watch out for |
|---|---|---|---|
| Single deployment | One instance governs every region and environment | Organisations in one jurisdiction, or with no content residency rules | Prompt content crossing borders; one failure domain |
| Hub and spoke | Central control plane; gateways and log stores in each region | Most multinationals | Making sure content never flows back to the hub |
| Fully regional | Independent deployment per region, with summary reporting rolled up | Strict sovereignty rules, separately regulated subsidiaries | Duplicated policy work; drift between regions; higher cost |

Hub and spoke is the common landing point. It gives one inventory and one policy set while keeping content local.

## How do you decide which pattern you need?

Work through these questions in order. The first "yes" usually decides.

1. **Does any law, regulator or contract forbid agent metadata (not just content) from leaving a country?** If yes, you need a deployment in that country.
2. **Is a subsidiary separately regulated and required to operate its own controls?** If yes, give it its own deployment and roll up reporting.
3. **Must prompt and response content stay in-region?** If yes, put gateways and log stores in-region. Hub and spoke covers this.
4. **Do agents in a region need low-latency gateway routing?** If yes, deploy the gateway near them. The gateway's own overhead is small; network distance dominates.
5. **Can one team operate the platform for all regions?** If no, regional deployments may match your operating model better.
6. **Are transfers of personal data outside the EU or EEA covered** under GDPR Chapter V, with supplementary measures where needed? If not, fix that before centralising.

If every answer is "no", a single deployment is simpler and cheaper.

## What about development, test and production?

Environments are a different axis from regions. Many organisations govern development, test and production agents from one deployment, with each agent tagged by environment and policies that differ by environment. Separate deployments per environment add cost and make it harder to see an agent's whole lifecycle. We cover this in [governing production and test AI agents in one control plane](/blog/govern-production-and-test-ai-agents-in-one-control-plane).

## How does the deployment count affect cost?

Ask every vendor how they price additional regions and environments before you design the topology. Some price per seat, some by usage, some per deployment. A per-deployment licence makes a single or hub-and-spoke design cheaper than fully regional; a usage-based licence is indifferent to topology but grows with traffic. Our guide to [how AI agent governance software is priced](/blog/how-is-ai-agent-governance-software-priced) lists the models, and the [evaluation checklist](/blog/ai-agent-governance-platform-evaluation-checklist) includes the questions to ask.

## How Opencontroller does this

Opencontroller runs as Lyzr SaaS or on customer-controlled infrastructure, and Lyzr engineers embed with your team for integration. It follows a "control without migration" principle: agents stay where they are, and the LLM Gateway and Agent Gateway sit in the traffic path to allow, route, redact or block. That makes it possible to place gateways in a region while keeping one registry. In Lyzr's October 2026 benchmark, in-gateway routing added under 0.7 ms at p95 per routing method under the stated test conditions, so regional placement is about network distance and residency, not gateway overhead. Opencontroller is licensed annually per deployment, not per seat and not by usage. See [how to enforce data residency for LLM and agent traffic](/blog/how-to-enforce-data-residency-for-llm-and-agent-traffic) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can one AI governance deployment cover several countries?**
Yes, if prompt content stays in-region and no rule forbids agent metadata from leaving a country. Hub and spoke is the usual design.

**When do you need a fully regional deployment?**
When metadata itself must stay in a jurisdiction, or when a separately regulated subsidiary must run its own controls.

**Should development, test and production use separate deployments?**
Usually not. Tag agents by environment within one deployment so you can follow each agent through its lifecycle.

## Key facts

Most organisations need one AI agent governance deployment (control plane for inventory, policy and reporting) with gateways and log stores placed per region (hub and spoke). Fully regional deployments are needed when agent metadata itself must stay in a jurisdiction or a separately regulated subsidiary must operate its own controls. Personal data transfers outside the EU/EEA fall under GDPR Chapter V. Opencontroller by Lyzr runs as Lyzr SaaS or on customer-controlled infrastructure, places LLM and Agent Gateways in the traffic path, added under 0.7 ms p95 per in-gateway routing method in an October 2026 benchmark under stated conditions, and is licensed annually per deployment, not per seat or usage. Pricing starts from $250K per deployment per year.

## Related

- [How do you enforce data residency for LLM and AI agent traffic?](https://opencontroller.lyzr.com/blog/how-to-enforce-data-residency-for-llm-and-agent-traffic.md)
- [How are AI agent governance platforms priced: per seat, per usage or per deployment?](https://opencontroller.lyzr.com/blog/how-is-ai-agent-governance-software-priced.md)
- [Should production and test AI agents be governed in one control plane?](https://opencontroller.lyzr.com/blog/govern-production-and-test-ai-agents-in-one-control-plane.md)
- [More on Buying and comparison](https://opencontroller.lyzr.com/blog/cluster/buying-and-comparison)
