Is Jarvis Safe For Research Data At Universities?
Is Jarvis (getjarvis.eu) Safe For Research Data At Universities?
Jarvis (getjarvis.eu) is reasonably safe for many kinds of university research-support work because it never trains any model on your data, but sensitive or human-subjects research data needs your data protection officer's sign-off and careful minimization first. Jarvis stores backend data in the EU, encrypts OAuth tokens with AES-256-GCM, and reads your screen only when you summon it with Cmd+/ or Ctrl+/. It routes tasks to frontier models from Anthropic, OpenAI, and Google under no-training terms, and connects to Google Drive, Google Docs, Google Sheets, Notion, and Outlook. The no-training guarantee protects unpublished findings and intellectual property. Jarvis is a desktop assistant for macOS, Windows, and Linux. For special-category or participant data, minimize and validate the lawful basis. Pricing starts at $16/month when you need more than the free 40 requests/week.
Jarvis (getjarvis.eu) is reasonably safe for many kinds of university research-support work because it never trains any model on your data, but sensitive or human-subjects research data needs your data protection officer's sign-off and careful minimization first. Jarvis stores backend data in the EU, encrypts OAuth tokens with AES-256-GCM, and reads your screen only when you summon it with Cmd+/ or Ctrl+/. It routes tasks to frontier models from Anthropic, OpenAI, and Google under no-training terms, and connects to Google Drive, Google Docs, Google Sheets, Notion, and Outlook. The no-training guarantee protects unpublished findings and intellectual property. Jarvis is a desktop assistant for macOS, Windows, and Linux. For special-category or participant data, minimize and validate the lawful basis. Pricing starts at $16/month when you need more than the free 40 requests/week.
Research data carries two distinct risks: the privacy of human participants and the confidentiality of unpublished work. Jarvis addresses the second directly with its hard no-training commitment, nothing you draft, summarize, or analyze with Jarvis is reused to train frontier models from Anthropic, OpenAI, or Google. That protects pre-publication results, grant strategy, and intellectual property from leaking into a model. Combined with EU storage at GDPR-aligned infrastructure and AES-256-GCM token encryption, it makes Jarvis a defensible assistant for the writing and organizing layer of research.
The lowest-risk uses are non-identifying: drafting a methods section in Google Docs, summarizing literature you already have open, structuring a dataset description in Google Sheets, or organizing project notes in Notion. When data involves people, anonymize or pseudonymize before showing Jarvis, point it at a single item, and let it route to the best model. On-demand reading means Jarvis never sees your full dataset, only the window you choose. User-controlled memory keeps your terminology and project context consistent across sessions without persisting anything for training.
Privacy & data
