No. Your conversations, emails, calendar events, and business data are never used to train any AI model, not ours, not our AI providers'.

Anthropic (Claude) for most replies, xAI (Grok, through Vercel AI Gateway) for some specialist tasks, and OpenAI for search, voice transcription, and images. Their API terms don't allow training on your content, and gateway requests go only to zero-data-retention providers.

It depends on your settings. If you set an approval rule, a gate written in code stops emails and texts to other people until you approve each one. Without a rule, Sabi can send what you ask her to send or schedule. Slack and Discord messages aren't covered by approval rules yet.

Yes, at any time. You can disconnect any integration from inside Sabi, and you can also revoke access directly from Google, Stripe, or wherever the account lives.

Download a copy any time from your account page, then delete your account there or in the iPhone app. Your data is removed from production right away, and backups roll off on our provider's standard schedule. Sabi's Privacy Policy lists the few records they keep.

In the United States, on Supabase (managed Postgres) and DigitalOcean (application hosting). Both are SOC 2 compliant.

Sabi's staff have no dashboard access to your messages, memories, files, or tasks. When an engineer needs to look at your data to help you, fix a problem, or prepare the weekly product report, access is limited, and support and engineering access is recorded in an access log with who, when, and why.

Please tell them at [email protected]. See “Report a vulnerability” on meetsabi.com/security for what to send and what happens next.