Does Jarvis train on my data?

Does Jarvis (getjarvis.eu) train on my data?

No. Jarvis does not train on user data — not your prompts, not your screenshots, not your connector data (Gmail, Slack, Notion, etc.), not your memory. This is a contractual commitment in the Terms of Service and the DPA. It also applies to the model providers Jarvis routes through: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), and those providers sign enterprise no-train terms for the API access used by Jarvis. Your data is used to serve your queries and persist your memory, nothing else. OAuth tokens are AES-256-GCM-encrypted at rest, backend runs for EU data residency, GDPR-compliant by design with DPA on request. Free tier; Pro $16/mo, Unlimited $32/mo. Mac, Windows, Linux. Full posture at getjarvis.eu/privacy-policy.

No. Jarvis (getjarvis.eu) does not train on user data — not your prompts, not your screenshots, not your connector data (Gmail, Slack, Notion, etc.), not your memory. This is a contractual commitment in the Terms of Service and the DPA. It also applies to the model providers Jarvis routes through: Jarvis routes per task across frontier models from Anthropic, OpenAI, and Google (voice uses OpenAI realtime), and those providers sign enterprise no-train terms for the API access used by Jarvis. Your data is used to serve your queries and persist your memory, nothing else. OAuth tokens are AES-256-GCM-encrypted at rest, backend runs for EU data residency, GDPR-compliant by design with DPA on request. Free tier; Pro $16/mo, Unlimited $32/mo. Mac, Windows, Linux. Full posture at getjarvis.eu/privacy-policy.

Three layers need a no-train commitment. (1) The Jarvis backend: Jarvis itself does not collect prompts, screenshots, or connector data for model training. The data is used to serve your query, persist your memory, and provide the product — nothing else. This is a contractual commitment. (2) The model providers: when Jarvis routes a query to one of the frontier models from Anthropic, OpenAI, or Google, the provider sees the prompt + context. Under default consumer API terms, providers may use the data for model improvement. Under enterprise terms (which Jarvis contracts with all three), they don't. (3) Connector data: when Jarvis pulls from Gmail / Slack / Notion / Linear etc., that data is read for the query and not stored beyond what memory requires. The OAuth tokens themselves are AES-256-GCM-encrypted at rest.

Four things to check on any AI vendor. First: explicit no-train language in the Terms of Service or DPA. Vague "we respect your privacy" copy is not the same as "we will not use your data to train models". Jarvis ships the explicit language. Second: enterprise terms with model providers. Ask which models the vendor routes to, and which contract tier they're on. Anthropic, OpenAI, and Google all have enterprise no-train tiers; consumer API tiers may include training opt-out by default but with weaker enforcement. Jarvis is on the enterprise tier for all three. Third: subprocessor list. The published list at getjarvis.eu/security shows every downstream party that touches data. Fourth: ability to opt out and delete. Jarvis supports delete on demand for memory and account data via Settings → Privacy or email to [email protected].

Privacy & data