Lumnyaailumnya.ai

Security by design

Last updated: July 5, 2026

An AI that talks to your shoppers and sees your orders has to earn trust with mechanisms, not promises. These are the mechanisms.

Encryption everywhere

All data is encrypted in transit (TLS) and at rest (Google Cloud — Firestore and Cloud Storage encrypted by default).

Strict per-store isolation

Every data read and write passes through an access layer bound to the specific store — one tenant can never see another's data. AI agents never touch the database directly: they work through controlled per-store layers with a full action trail.

Attachments: private by default

Support screenshots live in a private bucket with no public access; viewing happens only through time-limited signed URLs.

Limits everywhere: the AI can't run away

  • Rate limits per visitor, session and store on every public endpoint
  • A daily AI-call ceiling per store + a merchant-controlled kill switch
  • Usage charges stop at the cap you approved — never a surprise bill
  • AI money-actions (e.g. discounts) are logged and reversible

Secrets and authentication

Operational secrets travel only in headers (never URLs) and are compared with constant-time algorithms. Merchant authentication is handled by Shopify (OAuth); we store no passwords.

Deletion you can prove

Uninstall triggers full data erasure (conversations, settings, tickets, files), with automatic retry on transient failure. Inactive shoppers' personal details expire automatically under retention windows.

Monitoring

Real-time error tracking (Sentry) and periodic internal security reviews with documented findings and remediations.

Responsible disclosure

Found a vulnerability? Write to contact@lumnya.ai (subject: Security). We respond promptly, we don't pursue good-faith research legally, and we credit researchers who want it.