Brainz privacy policy
Effective September 4, 2026.
Brainz is operated by Matthew Sinclair Flath in Saskatchewan, Canada. This policy explains the data handled by the Brainz service.
Data we handle
- Account and authentication data: OAuth client metadata, connection and session state, and credential or token material needed to authenticate requests. In the gateway database, account emails are stored as one-way digests, passwords as verifiers, and session and token identifiers as hashes; raw account emails and passwords are not stored there.
- User content: the memories, headlines, tags, metadata, tasks, reminders, project context, and other records a user explicitly asks Brainz to store or retrieve.
- Gateway security and reliability data: pseudonymous account, client, tenant, and connection digests; timestamps; status and error codes; and coarse duration and byte-count buckets. Gateway audit events do not contain memory bodies, prompts, passwords, access keys, or raw tokens.
How we use data
We use this data to authenticate the account, connect it to the correct private tenant, perform the requested Brainz tools, protect the service from abuse, diagnose failures, and recover from outages.
Processors and disclosures
The service is hosted on DigitalOcean. Off-host database dumps are encrypted by Brainz before storage with Microsoft OneDrive. Track C artifacts and filtered non-secret tenant configuration are mirrored separately to OneDrive and rely on its storage protections rather than Brainz database-backup encryption. Tool inputs and results pass through OpenAI when Brainz is used from ChatGPT or Codex and are also subject to the user's OpenAI settings and OpenAI's policies.
When remote semantic search is enabled, eligible text and search queries are sent to the tenant's configured embedding provider, which may be OpenAI or OpenRouter. Track A thought bodies marked high or sealed, or with unknown or conflicting sensitivity, are not sent for remote indexing; search queries can still be sent. Brainz does not sell personal data or use it for advertising.
Retention
User content and account records remain in the primary service until the user deletes the record or asks support to retire the account or tenant. Short-lived browser sessions, authorization requests, OAuth codes and access tokens expire and are removed in bounded cleanup cycles. Rate-limit windows older than 48 hours are removed.
Scheduled database backups on the service host are pruned after 30 days. Offsite database backups and separate artifact and configuration mirrors have no automatic deletion deadline under the current recovery policy. A live-service deletion may therefore remain in recovery copies until they are manually retired; recovery copies are used only for disaster recovery.
User choices
Users can review supported memories and records through Brainz tools, disconnect Brainz in ChatGPT or Codex, and ask support to correct or delete data or retire an account or tenant. Disconnecting stops new access but does not by itself delete stored content.
Do not store payment-card data, protected health information, government identifiers, passwords, API keys, or one-time codes in Brainz.
Contact
Privacy and deletion requests: matthew@saltmind.ca.