NavCoach Privacy Policy

Effective date: September 1, 2026

NavCoach is operated by Yu Zhang (“NavCoach,” “we,” “us,” or “our”). This Privacy Policy explains how NavCoach processes information when you use the NavCoach iPhone application and related cloud services.

1. Scope

NavCoach is a meeting-preparation and live AI-assistance service. It processes information you choose to provide so that it can prepare meeting context, transcribe speech, generate meeting preparation, and provide live coaching.

This policy applies to the NavCoach External Beta R0 service. It does not describe unrelated research systems or future NavCoach products.

2. Information NavCoach processes

Account and authentication information

NavCoach requires authentication through Sign in with Apple or Google sign-in. Depending on the provider and your choices, NavCoach, Apple or Google, and Supabase may process an account identifier, email address, and limited provider profile metadata. Supabase Auth stores the NavCoach account and provider identity.

The app stores the Supabase session in iOS secure storage so you can remain signed in and refresh your session. NavCoach does not ask for or store your Apple or Google password.

Pre-Meeting information

When you create or prepare for a meeting, NavCoach may process:

  • meeting titles, metadata, and messages;

  • typed meeting context and notes;

  • Voice Prep or Ramble microphone audio you intentionally record;

  • the transcript produced from Voice Prep audio;

  • edits or selections you make while reviewing that transcript;

  • submitted preparation text, AI-generated Prep, and Context information; and

  • identifiers used to associate those records with your authenticated account and meeting.

Voice Prep audio is streamed through NavCoach cloud infrastructure to AssemblyAI for speech-to-text processing. NavCoach presents the resulting transcript for review before you submit it. The current beta does not durably store raw Voice Prep audio in NavCoach-controlled storage. Reviewed and submitted transcript text, meeting context, Prep, and Context records are stored in the account-scoped NavCoach database.

Live Support information

Live Support begins only after you intentionally start it, complete the per-meeting participant-awareness confirmation, and grant iOS microphone permission if needed.

During Live Support, NavCoach may process:

  • Quick Voice Check and meeting microphone audio;

  • live transcript and speaker-related evidence created from that audio;

  • meeting context and bounded transcript excerpts selected by the live coaching system;

  • AI Watcher inputs and outputs used to decide whether coaching is appropriate;

  • AI Reader inputs and outputs used to generate coaching when legitimately triggered; and

  • session state, timing, provider status, and delivery acknowledgments.

Live and Quick Voice Check audio is streamed through NavCoach cloud infrastructure to Deepgram for speech-to-text processing. Bounded text and context may be sent to Anthropic and OpenAI to provide the Watcher and Reader functions.

The current beta does not durably store raw Live audio, a full Live transcript, or coaching history in NavCoach-controlled product storage. These items are held in process memory for the active session and delivery lifecycle.

Service and diagnostic information

NavCoach processes authentication, meeting, request, and session identifiers; timestamps; state transitions; latency; provider status; error categories; and limited device or network request metadata needed to operate, secure, and troubleshoot the beta.

NavCoach’s application logging contract is designed not to include full conversation bodies, raw audio, authentication tokens, or provider credentials. Google Cloud and Supabase generate infrastructure, database, authentication, security, billing, and audit metadata when they provide the service.

3. How NavCoach uses information

NavCoach uses information to:

  • authenticate users, maintain sessions, and protect account-scoped data;

  • create and manage meetings;

  • transcribe intentional Voice Prep and Live microphone input;

  • let users review Voice Prep transcripts before submission;

  • generate meeting titles, Prep, Context, and live coaching;

  • perform Quick Voice Check and operate the live-session workflow;

  • display active-listening and coaching state;

  • deliver and acknowledge coaching output;

  • diagnose errors, measure operational latency, secure the service, and prevent abuse;

  • delete meetings or accounts when requested through the app; and

  • meet legal, security, and platform obligations.

The one-time in-app information-use notice provides general transparency. It is not blanket consent to record future meetings and does not establish that another participant has consented. Before the first Live capture in each meeting, NavCoach asks the user to confirm that participants are aware and that the user has any permission required for that conversation. Users are responsible for using NavCoach only where permitted.

4. Microphone access and user control

NavCoach requests iOS microphone permission only when you intentionally invoke Voice Prep or Live Support. The app displays a visible listening or recording state while microphone capture is active.

You can choose not to use Voice Prep, deny or revoke microphone permission in iOS Settings, cancel the Live participant-awareness prompt, stop Voice Prep, or end Live Support. Denying microphone access prevents microphone-dependent features but does not prevent use of typed meeting context where available.

5. Service providers and disclosures

NavCoach sends information to the following service providers only as needed to operate the current beta:

  • Google Cloud provides Cloud Run application services, container hosting, Secret Manager, Cloud Logging, and limited Cloud Storage authority associated with historical self-reference objects.

  • Supabase provides authentication and the owner-scoped PostgreSQL product database.

  • AssemblyAI provides streaming Voice Prep and Ramble transcription.

  • Deepgram provides streaming Quick Voice Check and Live transcription.

  • OpenAI API provides Prep, Context, and the selected Live Reader.

  • Anthropic API provides meeting-title generation and the selected Live Watcher.

  • Apple or Google provides authentication, depending on the sign-in method you select.

NavCoach configures and uses these providers for the functions described in this policy and relies on their applicable agreements and privacy and security commitments. The current beta does not include an advertising SDK, cross-app tracking, a data-broker integration, a product-growth analytics SDK, a third-party crash SDK, or a billing provider.

NavCoach requires these service providers to process information only for the service functions described here and to provide protection of user information at least equivalent to the protections described in this policy and required by applicable law.

AssemblyAI

NavCoach uses AssemblyAI’s U.S. Streaming service. The production account is opted out of AssemblyAI’s Model Improvement Program. AssemblyAI states that, for opted-out accounts, Streaming audio and transcripts have zero data retention. AssemblyAI retains certain transcript metadata for logging and billing.

The account also displays a 30-day retention setting for asynchronous transcription. NavCoach does not use AssemblyAI asynchronous transcription in this product path, so that asynchronous setting does not govern NavCoach Voice Prep Streaming.

Deepgram

Every NavCoach Deepgram request sends mip_opt_out=true. Deepgram states that an opted-out request is excluded from its Model Improvement Partnership Program and that the request data is retained only for the time needed to process it. The project-wide default is not opted out, so the per-request setting is material and must remain enabled.

OpenAI API

NavCoach uses the OpenAI Responses API with store=false. The production organization has model-feedback sharing, evaluation and fine-tuning sharing, and input/output sharing disabled. OpenAI states that API data is not used to train or improve its models unless the organization explicitly opts in.

The account uses standard per-call API logging. Under OpenAI’s standard abuse-monitoring controls, logs may contain prompts, responses, and derived metadata for up to 30 days, subject to documented legal and safety exceptions. Encrypted prompt-cache state may remain for up to 24 hours. The store=false request setting prevents NavCoach from requesting stored Responses application state, but it is not Zero Data Retention. NavCoach does not claim OpenAI Zero Data Retention or Modified Abuse Monitoring.

NavCoach does not use OpenAI conversation, file, assistant, or vector-store objects in the proven beta path.

Anthropic API

The Anthropic organization uses its standard 30-day data-retention setting. Zero Data Retention is not enabled, user feedback is disabled, and the organization is not enrolled in the Development Partner Program.

Anthropic states that commercial API inputs and outputs are not used for model training by default unless the customer explicitly provides qualifying feedback or otherwise opts in. Anthropic ordinarily deletes API inputs and outputs within 30 days. If its automated trust-and-safety systems flag content as a Usage Policy violation, Anthropic states that it may retain the flagged inputs and outputs for up to two years and associated trust-and-safety classification scores for up to seven years. Other documented exceptions apply for law, separately agreed arrangements, and features with different retention. NavCoach does not make an Anthropic Zero Data Retention claim.

Supabase

NavCoach product rows remain in the owner-scoped Supabase PostgreSQL database until the user deletes the applicable meeting or account because the beta has no automatic product-row expiration.

The current production project has no scheduled project backups, Point-in-Time Recovery is not enabled, and no Log Drain is configured. Under the project’s current service configuration, Supabase provides one day of accessible API and database log retention and one hour of Auth Audit Log retention.

6. Storage and retention

NavCoach’s proven retention posture is:

  • Supabase Auth account and provider metadata persist while the NavCoach account exists.

  • Submitted meeting context, messages, reviewed Voice Prep transcript text, Prep, model-run, and Context records persist in owner-scoped PostgreSQL tables until the user deletes the meeting or account.

  • Draft content may exist temporarily in device or server memory before submission.

  • NavCoach does not durably persist raw Voice Prep, Quick Voice Check, or Live audio.

  • Full Live transcripts and coaching output are process-local and are not stored as durable product-history rows.

  • AssemblyAI provides zero data retention for Streaming audio and transcripts on the opted-out production account, while retaining certain logging and billing metadata.

  • Deepgram retains opted-out request data only for the duration necessary to process the request.

  • OpenAI standard abuse-monitoring logs may retain prompts, responses, and derived metadata for up to 30 days, subject to documented exceptions. Encrypted prompt-cache state may persist for up to 24 hours.

  • Anthropic ordinarily deletes API inputs and outputs within 30 days. It documents up to two years for inputs and outputs flagged as Usage Policy violations and up to seven years for associated trust-and-safety classification scores, plus legal and other documented exceptions.

  • The current Supabase project has one day of accessible API and database log retention, one hour of Auth Audit Log retention, and no scheduled project backups or Point-in-Time Recovery.

  • Google Cloud’s default application-log bucket is configured for 30 days. Google’s required audit-log bucket is configured for 400 days. These logs are intended to contain operational metadata rather than conversation bodies.

  • Historical self-reference database rows and referenced Google Cloud Storage objects may exist for an account. Account deletion fails closed rather than silently leaving a referenced object behind.

Deleting information from NavCoach’s primary database does not itself guarantee immediate deletion from provider security logs, billing records, audit records, or information a provider must retain under law. Provider-held information expires under the applicable controls described above.

7. Your choices and controls

You can:

  • choose typed context instead of Voice Prep;

  • decide whether and when to start a microphone feature;

  • review Voice Prep transcript text before submitting it;

  • cancel the per-meeting Live participant-awareness confirmation without starting Live capture;

  • stop Voice Prep or end Live Support;

  • change microphone permission in iOS Settings;

  • sign out and clear the local NavCoach session;

  • delete an individual meeting; and

  • initiate permanent account deletion from the in-app Profile & settings screen.

Deleting or withholding information may make the corresponding feature unavailable.

8. Meeting and account deletion

Meeting deletion removes the authenticated owner’s meeting root and related database records through the database’s owner-scoped cascade behavior.

In-app account deletion calls an authenticated, caller-only database function. The proven flow deletes the caller’s meetings and related owner-scoped product rows, then deletes the Supabase Auth user. It does not allow anonymous or cross-owner deletion.

If a historical referenced cloud object has not been purged, deletion refuses to complete so NavCoach does not silently orphan the object. The user can retry after the object issue is resolved or contact support.

Account deletion clears the local NavCoach session and information-use acknowledgment. It does not delete the user’s Apple or Google account. Users can separately manage NavCoach’s connection in their Apple or Google account settings. Provider security, billing, audit, or legally required records may remain for the periods described above.

9. Security

NavCoach uses authenticated HTTPS and secure WebSocket connections, owner-scoped database access, server-side credentials held in Google Secret Manager, iOS secure session storage, bounded operational logging, and release-time secret scanning.

No security measure can guarantee absolute protection. Users should avoid providing information that is not needed for the meeting assistance they request.

10. Processing locations

The proven NavCoach beta application services run in Google Cloud us-west1, and the Supabase project is in us-west-2. NavCoach uses AssemblyAI’s standard U.S. Streaming endpoint.

Other service providers may process information in the United States or other locations under their applicable service terms and data-transfer arrangements.

11. Privacy requests and applicable rights

Depending on where you live, applicable law may provide rights to request access to, correction of, or deletion of personal information, or to object to or restrict certain processing. NavCoach will evaluate verified requests under applicable law and the technical limits described in this policy.

Users can initiate account deletion in the app. For other privacy requests or beta support, contact support@mynavcoach.com.

12. Age eligibility

The NavCoach External Beta is intended for adults who are at least 18 years old. NavCoach is not directed to children, and we do not knowingly collect personal information from anyone under 18.

If you believe someone under 18 has provided personal information to NavCoach, contact support@mynavcoach.com so we can investigate and take appropriate action, including deletion where appropriate.

13. Changes to this policy

NavCoach may update this policy when its data practices or legal obligations change. The published policy will identify its effective date.

If a future change materially affects the in-app information-use disclosure, NavCoach may require acknowledgment of a new disclosure version rather than treating NAVCOACH_AI_PRIVACY_NOTICE_V1 as acknowledgment of materially different practices.

14. Contact

Operator: Yu Zhang

Privacy and beta support: support@mynavcoach.com