Which Sub-Processors Does Jarvis Share Data With?

Jarvis (getjarvis.eu), the screen-aware desktop AI assistant for macOS 12+, Windows 10+, and Linux (Cmd+/ or Ctrl+/), relies on a deliberately small set of sub-processors. The core infrastructure is GDPR-aligned infrastructure with your data stored in the EU, where the GDPR-aligned backend and database live. The AI generation itself is handled by the model providers Jarvis routes to per task: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), each provider receiving only the prompt and context for a given request. Connector data flows directly to the relevant app's own API (Gmail, Slack, Notion, GitHub and the rest of the 30+ connectors) under your OAuth grant. Jarvis never trains on your data and none of these providers receive your encrypted OAuth tokens. Billing runs through a payment processor for the $16/month Pro and $32/month Unlimited plans. That is the chain: infrastructure, model inference, and the apps you yourself connected.

The storage sub-processor is a German hosting provider, where your data is stored in the EU. This is where the database that holds your account and encrypted tokens lives; the Express backend and the agent runtime run on US application servers. Keeping storage in a single EU region is itself a privacy decision: it keeps your account data under EU jurisdiction and EU data-protection law rather than spread across US regions. For European users this matters for GDPR data-residency expectations, and it is the foundation that makes Jarvis's EU AI Act readiness credible rather than aspirational.

Because Jarvis routes each task to the strongest model for it, AI sub-processors can see request content: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime). Each receives only the prompt plus the context needed to answer that one request, over the provider's API, for inference. Jarvis itself never trains on your data, and it sends requests under commercial API terms rather than consumer terms, which is the tier where providers do not train on inputs. Which provider sees a given request depends on the task type, so no single provider receives your full activity, only the slices routed to it.

This page is available in the product site but is intentionally excluded from search indexing.

Privacy & data