Version 1.2 Effective June 18, 2026 Updated June 26, 2026

Sub-processor Disclosure

We use sub-processors for cloud infrastructure, AI features, subscriptions, payments, stem separation, observability, notifications, and website contact-form delivery. This page lists each one, what data they receive, and their privacy policies.

ClefScape LLC (“ClefScape”) uses the third-party sub-processors listed below when providing the Praxis Saxophone service. We have data processing agreements with each sub-processor and require them to handle your data in accordance with applicable privacy law.

Last updated: June 26, 2026 — Version 1.2

Google Firebase (Authentication, Storage, Cloud Functions, Crashlytics, Analytics, Cloud Messaging)

Purpose: User authentication, user data storage, file storage, serverless backend processing, crash and error reporting, anonymized usage analytics, and push notification delivery.

Data processed:

  • Email address and user ID (authentication)
  • Account profile, library metadata, uploaded audio files, separated stems, practice recordings, session metrics, personal-lick content, Coach conversation history
  • Anonymous device ID, app version, OS version (analytics and crash reporting)
  • Stack traces and device state at time of crash (crash reports)
  • Push notification tokens and message payloads

Data location: Primarily United States, with multi-region defaults. EU/EEA data subject to Standard Contractual Clauses where required.

Retention: Crash reports retained for 90 days. Analytics data retained per Firebase defaults (up to 14 months for user-level data). User content retention follows our Privacy Policy and User Ownership Policy.

Privacy information: Firebase Privacy and Security

Google Vertex AI (Gemini)

Purpose: Powers the Praxis Coach — generating textual and voice feedback on your practice sessions.

Data processed:

  • Your Coach message text
  • Personalization context built from your practice data (sax type, skill level, weekly goal, recent practice history, current Workout of the Day, top active personal licks by title/key/tempo/status, session metrics)
  • Personal-lick source attribution is excluded

Data location: United States

Retention: Session context is not retained by Vertex AI after the API call completes. Your data is not used to train Google’s foundation models.

Privacy information: Google Cloud Privacy Notice

Google Lyria (via Vertex AI)

Purpose: Generates custom AI backing tracks at the tempo, key, and style you specify.

Data processed:

  • Your prompt text describing the backing track you want
  • Tempo, key, and style preferences

Data location: United States

Retention: Per Google Vertex AI terms. Generated audio is delivered to ClefScape and stored in your private library; the original generation request is not retained by Vertex AI.

Note: Backing tracks are AI-generated instrumental accompaniments and do not reproduce copyrighted recordings.

Privacy information: Google Cloud Privacy Notice

LALAL.AI

Purpose: Stem separation of user-uploaded audio — separating an uploaded track into its instrument parts (e.g., a “lead sax” stem and a “band” stem).

Data processed:

  • The audio file you upload, transmitted only at the time of processing

Data location: As described in LALAL.AI’s privacy policy.

Retention: Uploaded audio is retained only as long as needed to deliver the separated stems back to you, per LALAL.AI’s terms (typically 24-48 hours post-processing). ClefScape stores the resulting stems in your private library.

Privacy information: LALAL.AI Privacy Policy

Note: We strongly recommend uploading only audio you own or have licensed. Do not upload copyrighted recordings you do not have rights to.

RevenueCat

Purpose: Subscription and entitlement management — tracking which tier (Free, Pro, Studio, Educator) your account is on.

Data processed:

  • Anonymous app user identifier
  • Subscription product identifiers
  • Purchase status and entitlement state
  • App Store / Google Play receipt validation tokens

Data location: United States

Retention: Per RevenueCat’s data retention policy. ClefScape receives only entitlement status; payment card data is never shared with us.

Privacy information: RevenueCat Privacy Policy

Apple (In-App Purchase processing)

Purpose: Processing subscription payments and in-app purchase transactions on iOS.

Data processed:

  • Apple ID identifier (handled by Apple — not received by ClefScape)
  • Purchase transaction tokens
  • Subscription state

Data location: Per Apple’s terms; varies by user region.

Privacy information: Apple Privacy — App Store

Google Play Store (In-App Purchase processing)

Purpose: Processing subscription payments and in-app purchase transactions on Android.

Data processed:

  • Google account identifier (handled by Google — not received by ClefScape)
  • Purchase transaction tokens
  • Subscription state

Data location: Per Google’s terms; varies by user region.

Privacy information: Google Privacy Policy

Arize Phoenix (AI Observability)

Purpose: Monitoring Praxis Coach response quality, debugging issues, and running automated quality evaluations (brevity, factual grounding, tool-use accuracy).

Data processed (default — redacted):

  • A one-way hashed identifier derived from your Firebase user ID (HMAC-SHA256, 12 characters — not reversible to your identity)
  • Session and request UUIDs
  • The prompt version we used
  • Tool names called in each Coach interaction
  • Personalization scope tier we selected
  • Your subscription tier, beta-tester status, app platform, and app version

Your raw Coach query text and the Coach’s response text are NOT sent to Phoenix in standard accounts.

Data processed (internal accounts only): Internal Praxis Saxophone accounts that have explicitly opted into raw-text logging may have their raw Coach queries and responses sent to Phoenix for quality investigation. Standard user accounts never have raw text sent.

Eval results and feedback: Deterministic quality eval scores and your thumbs-up / thumbs-down annotations are stored as span attributes alongside your hashed identifier.

Personal-lick exclusion: Personal-lick source attribution is excluded from any Phoenix payload, regardless of opt-in status.

Data location: United States

Training: Arize Phoenix is an observability platform. Arize does not train AI models on customer data; the service is monitoring, not model improvement.

Privacy information: Arize Trust Center

Web3Forms (Website contact-form delivery)

Purpose: Delivers messages submitted through the contact form on our website (clefscape.studio) to our support inbox. This applies only to the marketing website, not the Praxis Saxophone app.

Data processed:

  • The name, email address, and message you enter in the contact form
  • Basic submission metadata (e.g., timestamp)

Data location: European Union.

Retention: Web3Forms forwards your submission to our email and retains it only briefly per its privacy policy (approximately 30 days), after which it is deleted from their systems. The message itself remains in our support inbox.

Training: Web3Forms does not use submitted data to train AI models.

Privacy information: Web3Forms Privacy Policy

What We Do Not Share

We do not share user data with:

  • Advertising networks or ad-targeting providers (we do not run ads in Praxis Saxophone).
  • Data brokers, analytics aggregators, or behavioral ad networks.
  • Other users — every user’s data is private to their account; there is no public or social surface.

If we add any sub-processor in the above categories in the future, we will update this page and notify users.

Updates to This List

We will update this page when we add, change, or remove sub-processors. Material changes will be reflected in the “last updated” date and version number above, and announced per our Privacy Policy.

Questions? Contact support@clefscape.com.