Backfrom Databricks’s own policies · verified 2026-09-02
Databricks: data residency & retention
Headquarters
🇺🇸 United States (US)
Where inference runs
US, EU, Other
Foundation Model APIs run from the customer's workspace region (published pay-per-token / provisioned-throughput region lists across US, EU and other geos) and are a 'Databricks Designated Service' using Databricks Geos for residency. Caveat from their compliance docs: 'Databricks might process your data outside of the region and cloud provider where your data originated', and non-US/EU workspaces must enable cross-Geo data processing.
Region pinning
Yes, self-serve
Residency follows the workspace region/Geo the customer chooses; US and EU Geos process in-Geo, other regions require cross-Geo processing enablement.
Prompt retention
Retained · up to 30 days
Databricks may temporarily process and store inputs/outputs to prevent, detect and mitigate abuse — isolated per customer, stored in the same region as the workspace for up to 30 days, accessible only for security/abuse response.
Trains on API data
No
Databricks docs: for all paid accounts, Model Serving does not use inputs submitted to the service or its outputs to train models or improve Databricks services.
Zero data retention
Not offered
No zero-data-retention option disclosed; the 30-day abuse-monitoring store is not documented as waivable.
Compiled from Databricks’s published privacy policy, terms, and documentation on 2026-09-02. “Not disclosed” means Databricks publishes nothing on the point. Policies change without notice, so verify against the sources above before relying on this for compliance decisions.