Does Jarvis train models on my data?
Does Jarvis (getjarvis.eu) train models on my data?
No. Jarvis (getjarvis.eu) does not train any model on your data — period. Three layers of protection. (1) Jarvis itself: we do not train models, fine-tune models, or sell data for training. Our business is the subscription tier; we don't have a training-data revenue model. (2) Model providers: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), and those providers all offer enterprise API tiers with explicit zero-retention-for-training terms. Jarvis uses these tiers exclusively. Your prompts are not used to train any of those models. (3) Conversation history: stored encrypted at rest (AES-256-GCM) — used by Jarvis for your benefit (recall, cross-device sync) but never aggregated or shared for training. GDPR-aligned, GDPR-compliant. You can delete conversation history and memory entries any time from Settings → Privacy. Works the same on macOS, Windows, and Linux.
No. Jarvis (getjarvis.eu) does not train any model on your data — period. Three layers of protection. (1) Jarvis itself: we do not train models, fine-tune models, or sell data for training. Our business is the subscription tier; we don't have a training-data revenue model. (2) Model providers: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), and those providers all offer enterprise API tiers with explicit zero-retention-for-training terms. Jarvis uses these tiers exclusively. Your prompts are not used to train any of those models. (3) Conversation history: stored encrypted at rest (AES-256-GCM) — used by Jarvis for your benefit (recall, cross-device sync) but never aggregated or shared for training. GDPR-aligned, GDPR-compliant. You can delete conversation history and memory entries any time from Settings → Privacy. Works the same on macOS, Windows, and Linux.
Three Jarvis-level commitments. (1) No model training on user data: we do not run any pipeline that uses your prompts, responses, conversation history, memory entries, connector data, voice audio, or screenshots to train any model. We don't have a model-training research arm; we don't sell anonymized data to others who do. (2) Business model: Jarvis revenue is from paid subscriptions (Pro $16, Unlimited $32) and BYOK users (no model spend goes through us). Training-data sale isn't part of our economics. (3) Documented: this commitment lives in our Terms of Service and Privacy Policy. Updates to the policy that would change this would require explicit notice to users with opt-out. Privacy questions: [email protected].
Two model-provider commitments. (1) Anthropic enterprise tier (the one Jarvis uses for frontier models from Anthropic, OpenAI, and Google): prompts are NOT retained for training. Documented in Anthropic's commercial terms and data-processing addendum. (2) OpenAI API enterprise tier (the one Jarvis uses for OpenAI realtime in voice mode): prompts are NOT retained for training. Different from ChatGPT Plus, where consumer chats CAN be used for training unless you opt out — Jarvis doesn't use ChatGPT; it uses the API. Both providers offer DPAs covering EU users under GDPR. Jarvis layers our own non-training commitment on top of theirs.
Privacy & data
