Update, October 14, 2026: The switch applies to new requests. Requests already running can finish outside the EU, and enabling EU inference increases credit use.
Lovable’s EU inference setting lets Enterprise workspaces route supported AI model requests to designated European endpoints. Inference is the processing that turns a prompt into a model’s reply. This setting controls where that happens, separately from where your app stores its database.
What stays in the EU
While you build, the setting covers project chat, background tasks, security scans, and image generation. It also covers end users’ requests through the built-in Lovable AI service in published apps.
If no permitted endpoint can serve a covered request, it fails instead of being sent elsewhere. Lovable’s documentation gives the message “EU AI capacity is temporarily unavailable.” An AI feature in your published app can therefore stop working during an outage rather than use a model outside the permitted region.
Disabled features and tools outside the boundary
Video generation, dictation, voice conversations, and certain image editing features are unavailable with the setting on.
Some tools the agent uses still process data in the United States: web search through Exa, reading web pages through Firecrawl, and browser automation through Browserbase. Turning on EU inference does not move those tools’ processing to Europe.
Who can turn it on
Workspace admins or owners can enable it under Workspace settings, then Security, then Privacy & security. Changes apply to new requests and can take about 20 minutes. A request already running finishes where it started, which can be outside the EU. You must wait an hour between changes.
Lovable also says EU inference uses more credits. A workspace requiring European model processing therefore has both a narrower feature set and a higher credit cost.
End of article