Security
AZAI Hub reads sensitive tickets, code, and data. Here is how it handles them.
- Each customer runs on a separate deployment and database. No shared tenancy.
- Retention and deletion are set per deployment in the contract.
- Single sign-on through Microsoft Entra.
- Short-lived sessions with refresh-token rotation.
- Login rate limits and account lockout.
- Role-based admin and per-feature permissions.
- The Hub reads by default. Queries are checked as read-only before anyone runs them.
- Database changes arrive as drafts with preview, verification, and rollback.
- Customer replies are drafts until a person sends them.
- Runs on your own Anthropic or AWS Bedrock account, so model calls fall under your agreement with that provider.
- Each run sends the ticket, code excerpts, and query results it works on to Claude.
- Neither the Anthropic API nor AWS Bedrock trains models on your prompts or outputs by default.
- Every investigation runs inside a fixed tool budget.
- Signed webhooks with timestamp and replay checks.
- Telemetry scrubbed of sensitive fields.
- Rate limits on live streams and request size limits.
- Strict security headers: CSP, HSTS, and frame protection.
Need a security review or a questionnaire filled in? Bring it to the demo.
Book a demo