AI Transparency

Defined terms carry the meanings given in BeDeeper's shared legal definitions unless expressly redefined here.

---

In plain language

This is the honest, member-facing account of how The Lantern actually works inside BeDeeper — what artificial intelligence does, what leaves your sanctuary when it does, and what it will not do. The legal disclaimer sits alongside it in AI Disclaimer. BeDeeper's underlying commitments about your writing are in Content Ownership. Data handling in general is in Privacy Policy.

---

1. What The Lantern is

The Lantern is a compassionate companion for reflection, powered by artificial intelligence. It lives inside The Deep. It is not a therapist, counselor, physician, lawyer, financial advisor, coach, spiritual director, or emergency service. It is not the sanctuary itself — the sanctuary is BeDeeper. The Lantern is one voice within it.

BeDeeper's product design constrains The Lantern with something called the Lantern Covenant: it notices before it suggests, asks before it answers, holds patterns loosely, never owns meaning, and grows quieter as you grow wiser. The Covenant is why The Lantern feels the way it does.

2. Where AI is used

AI is used in specific places, not everywhere.

• The Deep. When you write to The Lantern, an AI model produces the reply. This is the primary surface where you talk to The Lantern.
• Bounded companions. In some Passage rooms and Compass contexts, The Lantern can be invited into the conversation with your explicit action. Compass maps are only used as context for The Lantern when you turn that on for a specific map.
• Background reflection. After a Lantern turn, BeDeeper may quietly run a smaller AI model over the turn to extract signals — patterns, themes, recurring language — that help The Lantern remember with you and help the sanctuary respond more thoughtfully in later visits. This runs on message content that is already part of your account; it does not create new outbound sharing beyond what section 5 describes.
• Optional journal summaries. If you ask for a summary of a Living Journal entry, a smaller AI model produces one. This is opt-in per entry.

3. When AI is used

AI is used when you invoke it. You choose to write to The Lantern. You choose to turn on a Compass map as context. You choose to ask for a summary. The Living Journal itself is a private writing space; entries you write there are not sent to an AI model automatically. Background reflection over a Lantern turn runs because you have already had a Lantern turn.

4. Human responsibility

You decide what to write. You decide what to keep. You decide what is true for you. The Lantern illuminates. It does not define. When it offers a reframing or names a pattern, treat it as an invitation rather than an answer — that is how it is designed. Meaning belongs to you.

5. What leaves the sanctuary, and to whom

When you send a message to The Lantern in The Deep:

• The message you just wrote is sent to the language-model provider named as a subprocessor in BeDeeper's Legal Foundation. At the time of writing, that provider is OpenAI.
• A bounded amount of surrounding context is sent along with the message so The Lantern can respond in a way that remembers you. The current default context includes: the accumulated messages of the current session (so a conversation flows), a limited number of recent Living Journal entries as short excerpts, a limited number of long-term memories that BeDeeper has extracted for you, your growth profile, weekly reflection signals, recent patterns and observations, Artifacts you have accepted, the current Passage context, and canonical BeDeeper teaching content where it has been enabled.
• What is not sent: the entirety of your Living Journal, unrelated Compass content that you have not explicitly enabled as Lantern context, another member's data (BeDeeper cannot access another member's data for you), or your payment information.

Background reflection over a Lantern turn also runs on message content and produces structured signals that stay inside BeDeeper.

The bounded excerpts BeDeeper sends to the model provider are what the provider needs to answer you. They are not sent to the provider for training. See section 8 and Content Ownership.

6. Accuracy limitations

Language models make mistakes. They can:

• Get facts wrong.
• Reproduce a pattern from their training that does not match your situation.
• Sound confident about something they should be uncertain about.
• Miss context that was obvious to you.

The Lantern is designed to offer reflection rather than assert facts, and to hold patterns as tentative invitations rather than conclusions. Even so, treat what it says as a companion's perspective, not a verdict.

Where The Lantern says something you know is wrong, wrong for you, or unhelpful, it is fine to disregard it, correct it, or ask again. The Lantern will not take offense.

7. Reflection, not advice

BeDeeper is a sanctuary for reflection. The Lantern is not:

• Not medical. It does not diagnose, prescribe, treat, or predict clinical outcomes. See Wellness Disclaimer.
• Not legal. It does not give legal advice. Consult a lawyer for legal decisions.
• Not financial. It does not give financial advice. Consult a qualified advisor for financial decisions.
• Not a crisis service. If you are in danger or crisis, please contact a real crisis service. Resources are listed in Wellness Disclaimer.
• Not a dream interpreter. By design, The Lantern does not interpret dreams. It may help you notice recurring symbols, emotions, or themes over time — noticing, not decoding.
• Not a fortune teller. It does not predict futures, omens, or prophecies.
• Not the last word. It does not assign your purpose, your Discovery, or "the lesson." Those are yours.

The legal expression of these limits is in AI Disclaimer.

8. Data handling

Three commitments sit under this section, in full text in Content Ownership § 3:

• BeDeeper does not sell member writing.
• BeDeeper does not send member writing to any third party for the purpose of training AI models.
• BeDeeper does not authorize its language-model provider — or any other subprocessor — to use member writing to train AI models.

The bounded context sent to the model provider so The Lantern can reply to you is used to produce that reply. BeDeeper's contractual and account posture with the provider is intended to reflect the third commitment above. Where BeDeeper cannot fully verify a provider's internal behavior against contract, that limitation is disclosed honestly in Content Ownership § 3. Confirmation of the current provider-side posture is tracked in BeDeeper's launch legal checklist § F.2.

Where BeDeeper introduces or changes an integration that could affect this posture — including provider-side retention or training toggles — this document and BeDeeper's Legal Foundation are updated first.

Storage of your writing inside BeDeeper — where journal entries and Lantern conversations live, and how they are protected — is described in Privacy Policy and Security & Data Protection.

9. Model providers may change over time

The Lantern's model provider may change. At the time of writing, dialogue is served by an OpenAI model and background reflection by a smaller OpenAI model. If BeDeeper adds, replaces, or removes a provider, the subprocessor list in BeDeeper's Legal Foundation is updated first and this document follows. The core commitments in Content Ownership — no sale, no third-party AI training — do not change with a provider change; they bind BeDeeper regardless of provider.

10. Continuous improvement

BeDeeper may adjust The Lantern's prompting, the Lantern Covenant, the balance of Passage rooms, safety filters, and how context is assembled without individual notice. These adjustments are ordinary product care — they can happen at any time. They do not release BeDeeper from the commitments in Content Ownership, the disclosures in this document, or the disclaimers in AI Disclaimer and Wellness Disclaimer.

Where an adjustment changes what you can expect from The Lantern in a way that meaningfully affects your rights — for example, a new default context source or a new subprocessor — BeDeeper updates the relevant document in this legal package and, where required, notifies you as described in Terms of Service.

11. Choice

You may:

• Use BeDeeper without ever opening The Deep. Passages can be walked without invoking The Lantern; the Living Journal is a private writing space.
• Turn off Lantern context for individual Compass maps by not enabling that setting.
• Delete a Lantern conversation from your Journey. Deletion effects are described in Account Termination.
• Close your account, as described in Account Termination.

12. Contact

Questions about how AI is used at BeDeeper may be sent to hello@bedeeper.com. Privacy questions may be sent to the privacy contact.