A one-page summary of how we run NewsroomCloud. It describes what is in place today — nothing more.
Hosting and regions
Application, database, search, queue, identity and media storage run self-hosted on virtualised servers we operate. The hosting provider and data-centre region are listed on the sub-processors page.
Article narration uses a self-hosted speech engine where configured, with Google Cloud Text-to-Speech as fallback. AI suggestions use the Google Gemini API, only when enabled.
Cloudflare provides CDN, DNS, attack protection and form bot checks (Turnstile) in front of the platform.
Encryption
All traffic to our sites and the panel is served over HTTPS (TLS) at the edge (Cloudflare).
Credentials saved through the settings panel (for example SMTP passwords, payment-provider and API secrets) and two-factor secrets are encrypted at the application layer with AES-256-GCM. The master key is kept outside the database and its backups.
Not encrypted yet: credentials stored in a tenant’s integration settings, and settings saved before August 2026 until they are saved again or migrated. These values are masked in every API response; moving them to the same encryption is in progress.
We do not currently claim disk or volume encryption on the servers, and database backups are not separately encrypted.
Card payments are handled by Stripe Checkout; card data never reaches our servers.
Access control
Sign-in runs through our self-operated Keycloak identity server.
Two-factor sign-in (TOTP) is available to every account but is not enforced yet, including for our own staff. Repeated failed sign-ins trigger brute-force protection.
Roles (author, editor, moderator, admin and more) map to capabilities that the API checks on every request.
Administrative actions are written to a per-tenant audit log (who, what, when).
Tenant isolation
Each customer’s data is separated by tenant in the database layer: scoped queries get the tenant injected automatically, and a query without tenant context fails instead of returning other customers’ data.
An automated cross-tenant test suite (articles, users, media, comments, billing, webhooks, API keys, ads, audit) runs against a real PostgreSQL database.
Backups and restore
A nightly PostgreSQL backup (pg_dump) is written to self-hosted S3-compatible object storage and kept for 30 days.
Not covered yet: an offsite copy (backups sit on the same host as the database), media files and the identity database.
Last documented restore drill: 2026-10-17 (run manually). The newest backup restored into an isolated database with 0 errors in about 2 minutes. Full service recovery has not been timed end to end, and scheduled weekly drills are prepared but not active yet.
Incident response
Alerts, rollback and recovery follow written runbooks; errors are tracked in our self-hosted Sentry.
We notify affected customers of a personal data breach without undue delay, as set out in the data processing agreement.
Current availability is shown on our status page.
We do not currently hold a security certification (such as ISO 27001 or SOC 2).