Privacy Policy

Local-only by default (no account required)

You can use Smidge in a zero-account, local-only mode. In this mode no account is required and your coaching data stays on your device. Your information is only associated with an account, and only synced off your device, if and when you choose to sign in.

Information we collect

Depending on how you use Smidge, we handle the following:

• Microphone audio and speech transcripts. When you talk to your coach by voice, microphone audio is captured and converted to text. Speech-to-text is performed by AssemblyAI, our transcription provider.
• Conversation text. The messages in your coaching conversation are processed by our AI partners to generate responses: Anthropic produces the coaching replies, and OpenAI provides text-to-speech (spoken voice output).
• Account identifiers. If you create an account, we use Supabase to manage authentication. This involves your email address, or the identifier provided by Apple or Google if you use their sign-in.
• Locally stored coaching data. Your journeys, goals, and coaching history are stored on your device. Local at-rest data is encrypted on device.

Optional, consent-gated data

Certain categories of processing are entirely optional. Every category below is OFF by default. Smidge works — with reduced precision in some cases — when none of them are enabled, and you may revoke any of them at any time in the app.

• Passive sensing. OS-level health data (such as sleep and heart-rate variability), calendar density, geolocation, and screen time, used to improve coaching precision.
• On-device voice analysis. On-device analysis of vocal qualities such as prosody, respiratory cadence, response latency, and speech fragmentation. Raw audio stays on your device.
• Remote voice analysis. A de-identified feature vector derived from your voice may be sent off-device for analysis. Raw audio never leaves your device — only de-identified features.
• Accountability partner. Sending success-only messages to a partner you configure, via SMS or email.
• Cross-user learning. Allowing your data to participate in privacy-bounded cross-user learning. At launch, no sub-categories of this are enabled.

How we use information

We use the information above to provide and operate the coaching experience: transcribing your speech, generating coaching responses, speaking replies aloud, maintaining your account if you have one, and keeping your coaching data available across your signed-in devices.

What we do NOT do

• We do not show advertising.
• We do not sell your personal data.
• We do not track you across other apps or websites.
• We do not embed third-party analytics or advertising SDKs in the mobile app.

Third parties who process your data

We share data with service providers only to deliver the product: AssemblyAI (speech-to-text), Anthropic (coaching responses), OpenAI (text-to-speech), and Supabase (authentication and account storage). Where you sign in with a third party, Apple or Google provides your sign-in identifier.

Retention and deletion

You can delete your account from within the app. Deletion performs a hard delete: your account and all associated per-user data are removed in a cascading deletion. We retain only a hashed, no-PII tombstone recording that the deletion occurred, and de-identified clinical-safety audit records (with your identifiers removed) that we are required to keep for safety governance. These retained records cannot be used to identify you.

Withdrawing consent

You can revoke any optional consent category in the app at any time. When you revoke a category, we stop the related collection, and any in-flight processing that cannot be interrupted is discarded rather than retained. Revoking a category may reduce coaching precision but Smidge continues to function.

Children

Smidge is not directed to children under 13, and we do not knowingly collect personal information from children under 13.

Contact

Questions about this policy or your data? Contact us at support@mysmidge.com. See also our Support page.