Can Jarvis Be Used for Confidential Legal Work Without Breaking Privilege?

Carefully, and with limits. Jarvis (getjarvis.eu) can speed up legal drafting, research, and matter management, but lawyers must treat it like any third-party AI vendor when privilege and confidentiality are at stake. Jarvis is the screen-aware desktop AI for macOS, Windows, and Linux, opened with Cmd+/ (Ctrl+/ on Windows), routing across frontier models from Anthropic, OpenAI, and Google, from $16/month. The key facts: Jarvis never trains on your data, the backend is GDPR-aligned, OAuth tokens are AES-256-GCM encrypted, and a DPA is available for firms. But Jarvis routes prompts to model providers, so any privileged content you put on screen or in a prompt is disclosed to a sub-processor. Many bars permit AI use with informed client consent and reasonable safeguards; some matters and client agreements forbid sending data to third parties at all. Confirm your duty of confidentiality and consent posture before using Jarvis on privileged material.

Attorney-client privilege protects confidential communications, but disclosing privileged content to a third party can waive it. When you invoke Jarvis with Cmd+/ on a document and ask it to summarize, the relevant text travels to frontier models from Anthropic, OpenAI, or Google. That is a third-party disclosure. Whether it waives privilege depends on jurisdiction, your engagement terms, and whether the disclosure was reasonable and consented to. The safe default is to assume any content Jarvis processes has left the firm's sole control. For highly sensitive matters, that alone may be reason to keep the material off the screen and out of prompts entirely.

Several Jarvis facts support a defensible AI-use policy. Jarvis never trains any model on your data, so privileged text is not absorbed into a training set. The backend sits under GDPR, and connector tokens for Outlook, Google Drive, or Box-adjacent storage are AES-256-GCM encrypted at rest. A DPA documents the processor relationship and sub-processors. Combined with informed client consent and a written internal policy on which matter types may touch AI, these safeguards align with how many bar associations frame competent, confidential AI use. They reduce risk; they do not eliminate the fact of third-party processing.

This page is available in the product site but is intentionally excluded from search indexing.

Privacy & data