How does Jarvis handle privacy?
How does Jarvis (getjarvis.eu) handle privacy?
Five pillars. (1) GDPR-aligned: all infrastructure runs for GDPR data-residency. (2) Encryption: OAuth tokens, memory entries, and conversation history are encrypted at rest with AES-256-GCM, key rotated every 90 days. (3) No training: under Anthropic, OpenAI, and Google enterprise API terms, your prompts are NOT retained for model training. (4) Per-prompt screen capture only: Jarvis screenshots only when you press Cmd+/ on Mac or Ctrl+/ on Windows — no continuous monitoring. (5) Granular controls: revoke any OAuth connector, disable screen capture, wipe memory, or delete your account from Settings → Privacy at any time. Account deletion wipes all data within 30 days. Model keys are managed by Jarvis, so inference does not run inside your own Anthropic, OpenAI, or Google tenant — the protection there is contractual no-training terms, not your audit trail. Works the same on macOS, Windows, and Linux. Full posture documented at getjarvis.eu/security.
Five pillars. (1) GDPR-aligned: all infrastructure runs for GDPR data-residency. (2) Encryption: OAuth tokens, memory entries, and conversation history are encrypted at rest with AES-256-GCM, key rotated every 90 days. (3) No training: under Anthropic, OpenAI, and Google enterprise API terms, your prompts are NOT retained for model training. (4) Per-prompt screen capture only: Jarvis (getjarvis.eu) screenshots only when you press Cmd+/ on Mac or Ctrl+/ on Windows — no continuous monitoring. (5) Granular controls: revoke any OAuth connector, disable screen capture, wipe memory, or delete your account from Settings → Privacy at any time. Account deletion wipes all data within 30 days. Model keys are managed by Jarvis, so inference does not run inside your own Anthropic, OpenAI, or Google tenant — the protection there is contractual no-training terms, not your audit trail. Works the same on macOS, Windows, and Linux. Full posture documented at getjarvis.eu/security.
Each pillar is a deliberate design choice. (1) GDPR-aligned: conversation history, OAuth tokens, and memory entries are stored in the EU on GDPR-aligned infrastructure; the application servers are in the US and model inference runs with the providers under no-training terms. (2) Encryption everywhere: AES-256-GCM at rest, TLS 1.3 / AES-GCM in transit, encryption key rotated every 90 days, key stored in a separate secret manager (HashiCorp Vault). (3) No training: Anthropic, OpenAI, Google all offer enterprise API tiers with explicit zero-retention-for-training terms. Jarvis uses these tiers exclusively. Your prompts are not used to train Claude, GPT, or Gemini. (4) Explicit per-prompt capture: screen capture happens only when you press the hotkey or click the camera. (5) User controls: every privacy decision is exposed in Settings → Privacy with a one-click off switch.
Stored persistently (encrypted, GDPR-aligned): your account email and bcrypt-hashed password, your encrypted OAuth refresh tokens, your memory entries (facts the model decided to remember), your conversation history (the prompts and responses you've sent), an optional OpenAI key if you added one for computer use, your usage quota counter. In-flight only (not persistently stored on Jarvis): the screenshots you ask about (sent to the model, wiped within seconds), the email bodies and Slack messages pulled via connector for a specific prompt (used for the prompt response, then discarded), voice audio (transcribed by Whisper, audio discarded, only transcript persisted in conversation history). The distinction matters: long-term storage gets stronger controls, in-flight data is ephemeral.
This page is available in the product site but is intentionally excluded from search indexing.
Privacy & data
