Security

Orishare runs on Google Cloud in us-central1 behind Cloudflare. This page describes what is built into the platform today; it is not a certification claim.

Tenancy

Every operational row carries an environment id and a runtime class. Every query is scoped to one environment; sandbox and production of the same project never share rows. Staging and production runtimes use separate database roles that can only see their own environments.

Credentials

Secret keys (osk_), publishable keys (opk_), personal access tokens (opat_) and browser sessions (oses_) are random, stored hashed, shown once, scoped, and revocable individually.
Publishable keys can only send events an editor has explicitly allowed from the client, for a customer holding a short-lived client token.
Sign-in uses Google Cloud Identity Platform; Orishare never stores passwords.
Webhooks are signed (HMAC-SHA256 over timestamp and body) with per-endpoint secrets that rotate with a 24-hour overlap.

Determinism and audit

Rules are immutable versions. The same event, customer state and ruleset always produce the same effects, and every evaluation keeps its plan and trace. Every configuration change writes an audit entry with actor, before and after. Staff access to customer organizations goes through an audited support console that requires a reason for every request.

Personal data

Customer attributes flagged as personal data are excluded from AI drafting context. Deleting a customer redacts attributes and events, removes aliases and keeps only the financial ledger history. A per-customer data export returns everything held about a person in one document. Sandbox data is retained 30 days, production 90 days in the hot store with history in an archive you can query.

AI

The model only ever drafts configuration for a human to review, simulate and publish. It never runs on the event path. Drafts are validated against your real events, currencies and rewards before you see them, and every generation is stored with the context it saw.

Reporting

Report vulnerabilities to [email protected]. We acknowledge within two business days.