EU AI deployment beyond the regional endpoint

Factory's EU deployment keeps session content and inference in Europe, while organization pointer records, user profiles, and billing data remain in the global US deployment. That allocation is the starting point for an EU AI deployment review because different categories of information follow different paths.

The regional deployment documentation describes a dedicated backend, database, and inference endpoints. Factory still operates that regional service. An organization that also requires customer-operated infrastructure must make a separate control-plane ownership decision.

EU AI deployment starts with data categories

Write the requirement against the information the agent handles. Repository content used in a prompt, tool results, session messages, authentication records, and billing records may have different owners and retention rules. Review each category with the team responsible for it.

Factory documents that EU organizations use EU-region inference endpoints. Models or providers unavailable in that region are hidden from the picker and rejected server-side. That is an enforceable product behavior, but it still leaves the organization responsible for deciding which providers meet its contractual and legal requirements.

Map integrations and telemetry separately. A ticket, document, or trace sent to another system has its own access, storage, and regional conditions. A regional agent deployment does not relocate a connector's source system or a customer's telemetry collector, and it does not rewrite their contracts.

For the pilot, use representative but approved content. Follow it from the repository through inference and into the retained session. Include a denied model selection in the test, then verify the configuration and observed result rather than relying on the label in the user interface.

EU AI deployment and customer ownership differ

Factory Private places the control plane in customer infrastructure. Organizations can deploy it in a VPC or on-premises environment, with airgapped configurations available. This introduces operating responsibilities for infrastructure, upgrades, model availability, and recovery.

Choose Factory-operated regional hosting when its documented data allocation and provider arrangements meet the requirement. Choose customer-controlled deployment when policy requires ownership of the control plane. A fully disconnected environment adds approved local inference and an offline dependency process. Neither regional placement nor customer ownership alone establishes compliance with a particular law or contract.

The resulting decision should name the data classes, approved regions, operators, model services, and exceptions. Attach the relevant contracts and test evidence. This gives security and procurement teams something more useful than a regional badge.

Revisit the record when enabling a connector, changing model providers, or adding telemetry content. Each can introduce a new data path even when the original deployment stays in the same region.

Further reading

• AI code data residency
• Self-hosted AI agents
• Factory data flows and privacy