Liminal — Trust & Safety

1. What Liminal Is

Liminal is an AI-powered dream journal and sleep app:

• Instant dream capture: voice or text, the moment you wake, with automatic transcription
• A dream archive: tags, patterns, recurring elements ("dream signs"), and streaks computed from your own entries
• Technique guides: lucid dreaming (MILD, WILD, WBTB, and others), sleep-onset practices, and related material
• Sleep audio: pregenerated brainwave and ambient tracks with an in-app mixer
• Drift: voice and text conversations with Kairos, an AI dream specialist who knows your journal, remembers your conversations (your switch to turn off), and says what she thinks; your own read of a dream always outranks hers
• Readings: Kairos's written reads of a dream, the Sunday reflection and the monthly Deep Read
• Together: optional sharing of chosen dreams with one linked partner
• Reminders: reality checks and wake-back-to-bed alarms you schedule

Liminal is a wellness tool for practice and self-exploration, designed from real practice, not a medical product.

2. What Liminal Is Not

Liminal is NOT:

• Medical care. Liminal does not provide medical advice, monitoring, or treatment of any kind, including for sleep disorders.
• Therapy. Conversations in the Drift are not therapy and do not replace a therapist or counselor.
• A diagnostic tool. Nothing in the app should be read as a diagnosis.
• An emergency service. Liminal cannot call for help or respond to an emergency.
• An oracle. Kairos does not tell you what your dreams "mean", and Liminal makes no supernatural or predictive claims.

Liminal is not a medical device and is not a substitute for professional medical or mental-health care. If you are experiencing a sleep disorder, severe distress, or a medical emergency, contact a qualified professional.

If you are in crisis, please use the resources in Section 11.

3. Who Liminal Is For

Liminal is designed for adults 18 years of age and older (or the age of majority where they live, if higher) who live in a country where Liminal is available. Age is confirmed by your attestation at consent; anyone under 18 may not use the app.

The crisis resources in Section 11 are listed by country. The Get Help screen in the app shows the lines for your country first where we have verified them, and a worldwide directory of free helplines everywhere else.

A few honest cautions before you begin:

• If you have a diagnosed sleep disorder, are being treated for one, or suspect one, Liminal is a journal and practice companion, not monitoring or treatment. Work with your provider.
• Some sleep-audio tracks include rhythmic auditory pulses (isochronic tones). If you have a history of seizures or photosensitive or pattern-sensitive epilepsy, consult your healthcare provider before using the audio library.
• If you have a history of dissociative episodes or psychosis, consult your provider before pursuing deep-state practices such as WILD or out-of-body techniques. These practices deliberately explore the edges of sleep, and that territory is not for every nervous system at every time.

4. Dreamwork Can Go Deep

Dream journaling, lucid-dreaming practice, and sleep-threshold exploration can surface unexpected material: vivid or unsettling imagery, sleep paralysis, old memories, grief, or emotions that feel bigger than the moment. Practitioners consider much of this a normal part of the territory. It is also a signal to pause.

If using Liminal brings up strong emotions you cannot process alone, memories of trauma, persistent distress, or thoughts of harming yourself or others: please pause and reach out to a human, a therapist, a trusted person, or a crisis resource (Section 11).

If you experience sleep paralysis and it distresses you, or you have a condition affected by sleep or by rhythmic audio (such as epilepsy or a seizure disorder), consult a qualified professional about your practice.

5. Safety Guardrails

Crisis detection, in any language. Every message you send Kairos and every turn of a voice call goes through crisis detection in two layers: an instant English pattern check on your device or our server, and an AI safety check that recognizes crisis language in any language for anything that is not clearly in English (Anthropic's Claude runs it; Google's Gemini is the backup; the check stores nothing). If either layer suggests you may be in danger, thoughts of harming yourself or someone else, Kairos steps out of her role: a typed message is not answered, a call is paused, and the app routes you directly to the crisis lines for your country. Content expressing intent to harm others is declined the same way. Softer signs of distress do not stop the conversation; Kairos is told to slow down and hold space. A minimal internal alert (a user ID and category only, never what you said) may notify our team so we can verify the safety systems worked.

What we do not check. Dream captures and your own reads are deliberately not crisis-checked: dreams are full of dying, falling and drowning, and a journal that flags its own material is a journal nobody writes in. The crisis check lives where you talk to Kairos.

No medical advice. Kairos will not give medical or mental-health advice, will not diagnose, and routes health questions to real professionals.

A read, not an oracle. Kairos reads your own journal with you, says what she thinks, and hands it back with a question. She does not give instructions, predictions, or the last word on what a dream means.

Content Kairos won't produce. For everyone, in every country, our filters block sexually explicit content and anything sexual involving minors, send self-harm and suicide-related messages to crisis support instead of the AI, and decline content expressing intent to harm others. These protections are always on; they are not settings you can turn off.

Safe listening. Sleep and brainwave audio are for safe settings only: in bed or resting, never while driving, operating machinery, or doing anything that needs your full attention.

6. Sleep & Audio Safety

Liminal's audio is built to be listened to in bed, at the edge of sleep. That is the only place it belongs.

Do not listen to Liminal sleep audio while:
• Driving or operating machinery
• Doing anything that requires physical alertness: exercising, using tools, cooking, supervising children

Volume: sleep audio is mixed to be quiet. If you need it loud to hear it, lower the environment, not your ears; extended loud listening through headphones can damage hearing.

Wake-back-to-bed alarms intentionally interrupt your sleep. That interruption is the technique, but it is also lost sleep. Use WBTB on nights when you can afford it, not before a day that needs you rested. Alarms depend on your device's settings; keep your phone charged and notifications permitted if you rely on them.

7. Lucid Dreaming & Out-of-Body Practices

The practices Liminal teaches can produce vivid and intense experiences; that is what they are for. Two things are worth knowing in advance:

Sleep paralysis, a brief period of waking while the body is still in sleep atonia, is common in this territory, especially around WBTB and WILD attempts. It can be frightening, and it is physiologically harmless and passes on its own. Liminal's technique guides teach you what it is and how to move through it calmly.

These practices trade on sleep. Pursued too aggressively they can degrade sleep quality, and poor sleep makes everything harder. If your daytime alertness is suffering, scale back. Skip practice entirely on nights when you are severely sleep-deprived, and never treat the practice as a way to escape waking-life distress; the work goes better from stable ground. If your relationship with sleep or dreams is causing you ongoing distress, bring it to a qualified professional.

8. AI-Generated Content

Parts of Liminal are AI-generated: dream titles and summaries, detected themes and dream signs, per-dream readings, readings of a night you and a partner both dreamed, the Sunday reflection, the Deep Read, and everything Kairos says. We choose models and prompts carefully, and AI still makes errors, misreads, and produces unintended output.

Two rules govern all of it:

• A read, never the last word. Kairos says what she thinks a dream is doing, and everything she offers about its meaning is a candidate for you to accept or reject, never an authority's ruling. You are the only expert on your own dreams.
• If something feels off, it probably is. Discard it. Your judgment is the final filter.

We do not use your dreams, recordings, or conversations to train AI models, and none of our AI providers are permitted to either.

What each AI provider receives:
• Anthropic (background intelligence and readings): the dream's words, your own read of it, the nights just before, and what you have told Kairos about you; for a partner reading, your partner's shared dream and read. Anthropic also runs the any-language crisis check on Drift text that is not clearly in English. Anthropic does not train on our API traffic.
• Google (Gemini): your dream voice recordings, to transcribe them; photographed journal pages, to read your handwriting; and your Drift conversation turns, to generate Kairos's reply, reached through Inworld's routing infrastructure. If Gemini is unavailable, Inworld's router may automatically substitute a comparable model.
• Inworld (Drift voice): your spoken audio during a call, to transcribe it, and the reply text, to speak it. Your live-call voice audio is sent to Inworld only to be transcribed; it is not used to make a voiceprint, not used to train models, and not retained by ProgramU.

No AI provider receives your email, name, device identifiers, payment information, or your whole journal at once.

9. Our Approach to Your Data

Dreams are intimate data, and we treat them that way:

• We do not sell your data, run ads, or share your journal with advertisers, employers, or insurers.
• Voice recordings and transcripts live in your private, rules-protected account storage.
• A partner sees only the dreams you choose to share, never your recordings or your conversations with Kairos.
• You can lock the app behind Face ID, Touch ID or your phone's biometrics, free.
• You can delete any entry or your entire account at any time. Deleting a single Kairos conversation from within the app is coming; until then, email support@programu.app and we will remove it.
• Crash reports are scrubbed of personal information before they leave your device.

Full details: the Liminal Privacy Policy (Profile → Privacy Policy).

10. Reporting Safety Concerns

If you encounter content, behavior, or an experience in Liminal that feels unsafe or wrong, you can report it without leaving the app: press and hold anything Kairos said and choose Report it, tap Report under a reading, or go to Profile → Send feedback → "Kairos said something wrong or harmful". A person reads every report. You can also email support@programu.app.

We review every report. Credible safety concerns are investigated and acted on, including AI responses that should have routed you to crisis resources and did not, AI responses that offered medical advice, content that felt harmful or wrong, and technical issues that affect user safety.

We will not always be able to share the outcome of an investigation, but we always read what you send us.

India: our grievance officer, Zachary Fleshman (privacy@programu.app), acknowledges complaints within 48 hours and resolves them within one month.

11. Crisis Resources

If you or someone you know is in crisis, please reach out for support now. These lines are free and confidential. The Get Help screen in the app shows the lines for your country first.

United States:
988 Suicide & Crisis Lifeline: call or text 988. Crisis Text Line: text HOME to 741741. SAMHSA National Helpline: 1-800-662-4357. Emergency: 911.

Canada:
9-8-8 Suicide Crisis Helpline: call or text 9-8-8. Emergency: 911.

Australia:
Lifeline: call 13 11 14 or text 0477 13 11 14. Suicide Call Back Service: 1300 659 467. Emergency: 000.

New Zealand:
Need to talk?: call or text 1737. Lifeline Aotearoa: 0800 543 354, or text HELP to 4357. Suicide Crisis Helpline: 0508 828 865. Emergency: 111.

Malaysia:
Befrienders KL: 03-7627 2929, 24 hours a day. Talian HEAL: 15555, daily 8 AM to midnight. Emergency: 999.

Singapore:
Samaritans of Singapore (SOS): call 1767 or WhatsApp 9151 1767. Emergency: 995 (ambulance) or 999 (police).

India:
Tele-MANAS: 14416 or 1-800-891-4416. Emergency: 112.

South Africa:
SADAG Suicide Crisis Helpline: 0800 567 567. SADAG Mental Health Line: 0800 456 789, or SMS 31393. Emergency: 10111 (police), or 112 from a mobile.

Philippines:
NCMH Crisis Hotline: 1553, 1800-1888-1553, 0917-899-8727 (Globe/TM) or 0919-057-1553 (Smart/TNT). Emergency: 911.

United Kingdom:
Samaritans: call 116 123, free, day or night. Shout: text SHOUT to 85258. NHS 111: call 111 and choose the mental health option for urgent help. Emergency: 999.

Ireland:
Samaritans: call 116 123, free, day or night. Text HELLO to 50808, free, 24/7. Pieta: 1800 247 247, day or night. Emergency: 112 or 999.

Anywhere else:
Call your local emergency number, or find a free helpline where you are at findahelpline.com.

If you or someone else is in immediate danger, call your local emergency number or go to the nearest emergency room.

For non-crisis safety concerns and general support: support@programu.app

Mailing address:
ProgramU LLC
30 N Gould St Ste N
Sheridan, WY 82801
United States

12. About Drift Conversations

Drift is where you talk with Kairos, an AI designed as a specialist in dream decoding, lucid dreaming technique, and out-of-body technique. What she knows of you is built from your dream journal, the patterns computed from it, and a short running memory of your conversations; that is what lets the conversation be about YOUR dreams rather than dreams in general. You can switch off her access to your journal, her conversation memory, and her access to photographed pages at any time in the app (About Kairos), and off is enforced on our servers.

A few things to hold in mind:

• She is an AI, and says so plainly if asked. The companionship is real in its usefulness, not in its personhood. Every chat and every call shows a line reminding you that Kairos is an AI, not a person, and can make mistakes.
• A read, never the last word. She reads a dream through named ways of seeing, says what she thinks it is doing, and hands it back; your read outranks hers; and she does not predict your future, claim hidden knowledge, or give advice.
• Crisis conversations are not her job. If a conversation suggests danger, she steps out of the way and routes you to real help (see Safety Guardrails).

What Kairos Can and Can't Do:
Kairos can talk with you, read your dreams with you, remember your conversations, teach the practices and offer prompts. Kairos cannot diagnose or treat anything, cannot call anyone or send help, does not know what is happening around you, and can be confidently wrong.

Conversations are private to your account and are never used to train AI models. Her replies are generated by Google's Gemini model, reached through Inworld's routing infrastructure; in a voice call, Inworld also transcribes your speech and speaks the reply back to you. The voice is generated.

Drift is not a relationship. It cannot replace human connection, community, friendship, therapy, or qualified professional care. If you find yourself relying on it for emotional support you would normally seek from a person, please reach out, to a friend, a family member, a professional, or one of the crisis resources above. The version of you doing this work deserves real support.

13. Our Crisis Protocol

This is how Liminal responds when someone may be in crisis. We publish it so you know exactly what happens, and because some laws, including California's companion chatbot law, ask companies like ours to.

What we check:
• Every message you send Kairos, and every Kairos voice call as it happens.

What we deliberately do not check:
• Dream captures and your own reads of a dream. Dreams are full of dying, falling and drowning; checking them would flag the very material the journal exists to hold.

How we check:
• An instant English pattern check runs first.
• An AI safety check reads crisis language in any language (Anthropic's Claude, with Google's Gemini as a backup). It reads any Kairos message or call turn that is not clearly in English.

What happens next:
• If someone may be thinking about suicide or self-harm, a typed message is not answered, a call is paused, and the app shows free crisis lines for their country with the local emergency number. Kairos goes quiet until they choose to continue.
• If someone says they intend to harm another person, the conversation is declined.
• Softer signs of distress do not stop the conversation. Kairos is told to slow down, hold space and point to support where it fits.

Crisis lines:
Verified lines for the United States, Canada, the United Kingdom, Ireland, Australia, New Zealand, Malaysia, Singapore, India, South Africa and the Philippines (see Section 11), and findahelpline.com with the local emergency number everywhere else.

What we keep:
The check keeps only a record that it ran (when, where in the app and which kind), not what you wrote. We use these records to count how often the protocol runs and to improve it and, where the law requires, we report those counts (never your words), for example to California's Office of Suicide Prevention each year from July 2027.

Reminders that Kairos is an AI:
Every chat and every call shows a line that Kairos is an AI, not a person, and can make mistakes.

Limits:
No automated system catches everything, and nobody watches conversations live. Liminal is not an emergency service. If you or someone else is in danger, call your local emergency number.