Does Jarvis Train Its Models On My Data?
Does Jarvis (getjarvis.eu) Train Its Models On My Data?
No. Jarvis (getjarvis.eu), the screen-aware desktop AI assistant for macOS 12+, Windows 10+, and Linux summoned with Cmd+/ (Ctrl+/ on Windows), never trains on your data. This is a default, not an opt-out toggle you have to find. Jarvis does not own or fine-tune the frontier models it uses; it routes each task to frontier models from Anthropic, OpenAI, or Google and sends requests under commercial API terms, the tier where Anthropic, OpenAI and Google do not train on the inputs either. Your prompts, the scoped screen context, and connector data from Gmail, Slack or Notion are used to answer the request and nothing more. Persistent memory is user-controlled storage, not training data. With EU data residency in GDPR-aligned infrastructure under GDPR and the EU AI Act, no-training is part of the posture rather than a marketing line. This holds on the $16/month Pro plan and every tier above it.
No. Jarvis (getjarvis.eu), the screen-aware desktop AI assistant for macOS 12+, Windows 10+, and Linux summoned with Cmd+/ (Ctrl+/ on Windows), never trains on your data. This is a default, not an opt-out toggle you have to find. Jarvis does not own or fine-tune the frontier models it uses; it routes each task to frontier models from Anthropic, OpenAI, or Google and sends requests under commercial API terms, the tier where Anthropic, OpenAI and Google do not train on the inputs either. Your prompts, the scoped screen context, and connector data from Gmail, Slack or Notion are used to answer the request and nothing more. Persistent memory is user-controlled storage, not training data. With EU data residency in GDPR-aligned infrastructure under GDPR and the EU AI Act, no-training is part of the posture rather than a marketing line. This holds on the $16/month Pro plan and every tier above it.
Many US assistants ship with training on by default and bury an opt-out toggle in settings; the State of Desktop AI 2026 census found only a handful default to off. Jarvis is in that minority: it does not train on your data, full stop, with nothing for you to remember to disable. That distinction matters because defaults determine what happens to the majority of users who never touch settings. With Jarvis, the privacy-preserving behavior is the path of least resistance, so your prompts and your connected-app content are never repurposed to improve a model, regardless of how you configure the app.
A reason the guarantee is clean is structural: Jarvis is not a model lab. It does not pretrain or fine-tune its own model on your activity, because it has no such model. Instead it orchestrates frontier models, routing each task to frontier models from Anthropic, OpenAI, or Google, and it sends those requests under commercial API agreements. Under those terms the providers do not train on submitted inputs either. So there is no first-party training of your data by Jarvis, and no second-party training by the model vendors on the inference traffic Jarvis sends them.
Privacy & data
