Ehrliche Offenlegung der Datenflüsse.

Grenzen

Was wir nicht tun.

• Sell your data.

  We have never sold customer data. We do not have a commercial relationship with data brokers, marketing networks, or affiliated companies that would receive it.

• Train AI on your data.

  Anthropic and OpenAI both default to no-train on the API tier. We never opt in to training. We are pursuing zero-data-retention enrollment with OpenAI to remove their default 30-day abuse-monitoring retention.

• Share data across customers.

  Every tenant-scoped table enforces Postgres Row Level Security via a session variable that follows every query. A query without a tenant context returns zero rows. The boundary is enforced at the database, not at the application layer.

• Read your meeting transcripts.

  When live meeting recording re-enables (paused per ADR-0068 pending the consent-capture flow), Recall.ai will capture the transcript on your behalf. Today: you upload the transcript directly via the meeting dashboard. Our server reads each transcript once to produce the recap email you requested, and the recap goes through cross-family supervisor review before send. No human at Ataski reads the transcript outside an incident-response investigation, which is itself audit-logged.

• Write to your tools.

  Every integration we ship is read-only by default. OAuth scopes are pinned to read-only at registration. A separate, audited approval flow gates any future write capability.

Unterauftragsverarbeiter

Anbieter, die Bytes Ihrer Daten speichern.

Jeder ist durch einen schriftlichen Auftragsverarbeitungsvertrag gebunden. Sie betreiben eine bestimmte Funktion, die Sie nutzen; Bytes gelangen sonst nirgendwohin.