Security
Legal work can involve privileged correspondence, PII, confidential documents, and draft strategy. Jarel is designed with controls that help teams review how sensitive AI work is handled.
Compliance roadmap
We show certification and compliance work with its actual status: supported, in progress, or planned. No badge-washing.
Security practices
Customer documents, prompts, and generated content are not used to train Jarel models. Provider processing follows configured no-training terms where available.
Sensitive legal materials are protected with encryption in transit and at rest. Advanced key-management options should be discussed for eligible enterprise deployments.
EU residency is the default public position. Additional region options should be presented only when enabled for the customer's workspace.
Workspace access is designed around authenticated users, authorization checks, and auditability. Enterprise identity controls are configured where available.
Reads, writes, and AI calls are logged for review where available. Enterprise audit controls are designed to support traceability across sensitive workflows.
Security testing, dependency review, and external assessment evidence should be shared according to its current status, not implied before it exists.
Transparency
We publish our subprocessor list and notify customers in advance of material changes.
Request full list| Vendor | Purpose | Region |
|---|---|---|
| Vercel | Hosting and edge | US / EU |
| Vercel AI Gateway | Model routing | US / EU |
| Anthropic | LLM provider | US |
| OpenAI | LLM and voice services | US |
| Supabase | Database and storage | US / EU / UK |
| Sentry | Error monitoring | US / EU |
We share our security posture, compliance roadmap, DPAs, and security materials with prospects under NDA where appropriate.