Skip to the explanation

Higher Self transparency

How Higher Self works

What the coach can use, what leaves the app, which model answers, and which controls are live.

Higher Self can make mistakes. It is not a doctor, therapist, emergency service, financial adviser, or substitute for professional care.

What happens when you use Higher Self

Higher Self uses the information needed to answer your request and provide optional personalization.

Current

Current behavior

  • Anthropic and OpenAI are approved Higher Self providers. The current production route is Anthropic.
  • Higher Self is available only inside an authenticated BeTheOne account.
  • Messages travel over encrypted network connections. Higher Self is not end-to-end encrypted.
  • Automated Higher Self email and Telegram coaching are disabled.
Optional feature

Your choices

  • You can use Higher Self without optional personalization.
  • You can control which optional information categories Higher Self may use.
  • You can inspect or delete saved Higher Self memories in the app.
  • For privacy, access, or deletion requests, contact [email protected].

The request path

From your question to a grounded reply

A simplified view of how a Higher Self request is processed.

  1. Current

    You ask

    Your message and recent conversation provide the immediate question. The app saves your message locally so the conversation can continue.

  2. Optional feature

    A small context envelope is built

    If personalization is enabled, Higher Self selects relevant allowed categories for the current conversation.

  3. Current

    The API verifies the request

    BeTheOne verifies that the request comes from an authenticated account before processing it.

  4. Current

    The active provider generates the reply

    The active provider receives your message and only the selected context needed for this request. Higher Self must not claim access to information that was not included.

  5. Optional feature

    You can see why

    A data-used receipt names the categories, source type, and freshness that influenced the reply without exposing sensitive values on the first view.

Data selection

Relevant to this conversation, not everything in your account

Personalization is on by default for signed-in accounts. Higher Self still selects only the categories needed for the current conversation. You can turn any category off at any time in Higher Self settings.

May be selected

  • Profile and onboarding answers
  • Your daily disciplines and completion state
  • XP, level, streak, and summaries of your own activity patterns
  • Recent emotions or check-ins you have allowed
  • Selected health signals you have allowed BeTheOne to read
  • Your creator profile and your own recent creator content
  • Recent Higher Self messages and selected memories
  • Privacy-safe accountability aggregates when the feature is enabled

Not included by default

  • Active goals, because they are not populated in the current Higher Self request
  • Your feed as general coaching context
  • A friend’s private posts, journals, chats, health data, emotions, or detailed activity
  • Another person’s private content without a clear purpose and valid consent from everyone affected
  • Any category you have turned off in Higher Self settings
  • Unrelated categories that are not needed for the current coaching intent

Context receipt

Used for this reply

The receipt names categories and provenance, not the private values themselves.

Discipline progress
User dataUpdated recently
Streak
User dataUpdated today
Coaching preference
Remembered preferenceSaved by you

It did not use your feed or private friend content.

Coach contract

Warm, candid, and accountable to the evidence

A normal reply should feel like a short conversation, not a lecture.

How it should reply

  • Reflect briefly, name one relevant truth, offer one concrete next move, then ask at most one natural question.
  • Use a more demanding tone only when accountability is relevant.
  • Make plans longer only when you ask for a plan.
  • Keep recommendations traceable to user data, a remembered preference, BeTheOne content, or a researched source.

What it must never do

  • Invent or inflate progress
  • Diagnose, prescribe, manipulate, shame, or threaten
  • Claim access to goals, friends, health, memories, or other data that were not included
  • Turn a vulnerable disclosure into a discipline judgment
Grounding example

Higher Self should base accountability on the information actually provided for the reply, not assumptions about you.

Models and providers

The provider is disclosed, not hidden behind “our AI”

01

Current production route

Anthropic is the only active production route. Anthropic and OpenAI are the only approved Higher Self providers.

02

OpenAI evaluation

OpenAI is approved as the only second-provider candidate, but it is not active in production and no evaluation winner has been selected.

03

Provider processing

Under standard terms, provider API content may be retained for up to 30 days, with limited safety, legal, abuse-prevention, and feedback exceptions.

  • BeTheOne does not enable provider use of Higher Self API data for model training.
  • BeTheOne does not use provider-managed conversation storage for Higher Self.
  • OpenAI Higher Self requests use store: false.
  • Anthropic remains active in production until the frozen evaluation and founder approval are complete.

Zero-data-retention and EU regional processing are optional future improvements, not current promises.

Anthropic: Read retention detailsOpenAI: Read retention details

Privacy and security

Encrypted in transit. Not end-to-end encrypted.

The server and approved model provider must process readable message text and allowed context to generate a reply. Calling that end-to-end encrypted would be inaccurate.

Verified controls

  • Encrypted network connections
  • Account authentication is required
  • Access to stored account data is restricted
  • Private conversation content is prohibited from product analytics

Important limits

  • Higher Self must process readable message text to generate a response, so it is not end-to-end encrypted.
  • AI providers may retain data under their terms and legal obligations.
  • Do not use Higher Self for emergencies or as a substitute for professional advice.

Retention

Different data needs different limits

Retention depends on where the data is processed. We avoid promising deletion timelines that have not been verified end to end.

DataCurrent disclosureControl or gate
Conversations on your deviceStored on your device.You can manage conversations in the app. Device and operating-system backups may follow separate retention rules.
Cloud chat backupPrivate, access-controlled storage may be used to preserve your conversation and preferences.For an access or deletion request, use the available in-app controls or contact [email protected]. We do not promise an unverified deletion deadline.
AI-provider processingProvider API content may be retained for up to 30 days under standard terms, with limited safety, legal, abuse-prevention, and feedback exceptions.BeTheOne does not enable API data for model training or provider-managed conversation storage. OpenAI Higher Self requests use store: false.
Operational request logsLimited operational metadata may be retained.Metadata only. Raw private message, journal, image, prompt, and reply content is prohibited.
Product analyticsEvents and aggregates only. Raw journal and chat text is prohibited.Retention follows the applicable Privacy Policy and configured analytics controls.

Proactive coaching

External coaching is off until each channel earns consent

Automated Higher Self email and Telegram coaching are currently disabled.

In-appPushEmail: offTelegram: off

Safety and support

AI guidance has limits

Higher Self can support reflection and next steps, but it cannot assess your full situation or guarantee that advice is correct.

Policy revision 2026-07-19-rc4, reviewed 2026-07-19.