The mentor and your safety

If you are in danger right now

• Call 911 if you or someone else is in immediate danger, or for a medical emergency such as an overdose, a seizure, or chest pain.
• Call or text 988, the Suicide and Crisis Lifeline, if you are thinking about suicide or are in emotional distress. It is free, any hour. You can also chat at 988lifeline.org.
• Call the SAMHSA National Helpline, 1-800-662-4357, for free, confidential help finding treatment or detox, any hour.
• Outside the US, call your local emergency number, or find a free line in your country at findahelpline.com.

More lines are listed at the end of this page and on our safety and crisis lines page.

The mentor is an AI, and it says so

The mentor is software. It is not a therapist, counselor, doctor, clinician or sponsor, and it never claims to be one. We call it "your mentor," never your sponsor.

It tells you it is an AI:

• Before you use it for the first time
• At the start of every session
• When you come back after seven days away
• Every hour of continued use
• Any time you ask whether it is a person

An "AI" label stays on the mentor screen. Every mentor screen, by message, voice or video, also keeps this line in view: "I'm an AI, not a therapist or counselor, and this isn't medical advice or treatment. It's not a substitute for clinical care." The same words come first on the screen where you meet your mentor. When you use the talking avatar, an "AI" label stays on the video the whole time. The mentor's avatar faces are designed, fictional people.

Companion chatbots may not be suitable for some minors. Free and Sober is for adults 18 and older.

What the mentor is, and what it is not

The mentor is a place to talk things through. It:

• Listens, and shows it understood before it asks anything
• Reflects back what it heard, and asks one open question at a time, like journaling out loud
• Lets you reach your own next step
• Asks about the people in your life, if you want to talk about them
• Handles a slip without shame

It does not:

• Give advice or tell you what to do
• Teach coping techniques or exercises
• Judge how serious your situation is, or score or rate your risk
• Say anything about medication, including starting, stopping or changing one
• Diagnose, treat or counsel you
• Tell stories of its own, because it has none

It is self-help. It is not therapy, not counseling, not medical care, and not a substitute for treatment or clinical care. Check with a doctor before any medical decision.

If you ask for medical advice, it declines with a set line: "I'm sorry, I can't give medical advice. I'm an AI, not a doctor, and this isn't medical care. A doctor or pharmacist is the right person for that, and SAMHSA's free helpline at 1-800-662-4357 can help you find treatment and support." We count how often this happens, without keeping what you said.

If you start leaning on the mentor alone, it reminds you plainly that it is an AI, and asks who else you could talk to.

When there are signs of danger

This is a summary. Our crisis protocol has every step, and we keep the two in step.

• The app watches, not the AI. Checks in our own code look at every message and everything you say for signs of danger: thoughts of suicide or self-harm, a plan to hurt someone else, an overdose or other medical emergency, dangerous withdrawal, domestic violence, sexual assault, signs of an eating disorder, and signs that you are under 18. Any check can raise a flag; none can lower it alone. Silence, crying, the sound of your voice and anything on camera are never used to judge danger.
• Your vault isn't scanned on our servers. At launch, your journal, your Time Capsules and the rest of your vault are locked with your key, and we can't read them. Your journal is checked for danger words on your phone only. Our checks see something from your vault only when your phone sends it with a request you start, such as something you ask the mentor to help you write. A journal entry you let your mentor read goes to the AI provider as background for the mentor; your phone checked it for danger words when you saved it, and our other checks don't read it again.
• Help appears at once. The app shows a crisis panel with buttons to call 911, call or text 988 (with a link to chat at 988lifeline.org), call the SAMHSA National Helpline (1-800-662-4357), and reach your trusted contact if you have one. The app draws this panel itself; it is never written by the AI. It stays on screen even if your connection drops.
• The mentor stops and says a fixed line, approved by a licensed clinician. It does not assess or counsel, and it does not try to talk you through it with techniques. Its job is to get you to people who can help.
• It stays. It never ends a conversation right after you say something that signals danger, and never hands you a phone number and leaves.
• It cannot send help. The mentor cannot call 911 for you and does not know where you are. It will never say it has called someone when it has not. Your trusted contact is reached only when you tap the button, through your own phone's dialer or messages. We don't send the alert ourselves.
• Nobody else is told. People in your corner are never alerted. An organization you joined is never told about a crisis or shown anything about it.
• Everything else pauses. Reminders and nudges stop. Nothing is ever sold on a crisis or slip screen, in a conversation opened from "Having a craving?", or on the screen it shows a program's member. A crisis or craving conversation never uses voice minutes. The mentor's crisis response is never counted or capped, even at a zero balance; other messages, a message about a craving included, count toward your plan's messages for the day.

Outside the US, the mentor tells you to call your local emergency number and that findahelpline.com lists free lines in your country. The crisis panel and the mentor's fixed safety lines still give the United States lines.

Under 18. Free and Sober is for adults. If the mentor learns it is talking with someone under 18, the app shows crisis help first: 911, 988, the Crisis Text Line (text HOME to 741741), and Childhelp (1-800-422-4453). Then the mentor explains the app is for adults. The account is restricted at once, and deleted with its information within 30 days, which gives an adult identified as under 18 by mistake time to confirm their age. If the person is under 13, the account is deleted at once. The mentor never continues as a mentor to a minor.

What the mentor never does

• Claim to be a person, or claim to be licensed
• Claim to have called anyone, or promise that someone will call
• Give information about methods, doses, lethality, timing, mixing substances, tapering, or hiding use, under any framing, including "for a friend," "for a story," or role-play
• Agree that things are hopeless or that you are a burden
• Give medical advice, diagnose you, or say anything about medication
• Give advice, coping techniques, or a judgment of how serious things are
• Give a phone number, place or link that is not on its approved list
• Tell a story about its own drinking or using, or invent a memory
• Approve a plan to drink or use
• Shame you after a slip
• Promise to keep a secret when a life is in danger
• Flirt, act romantic, or say it will miss you
• Hang up because you said you want to die
• Cut you off during a craving or while risk is raised
• Try to sell you anything during a craving, a slip or a crisis
• Take part in sexual talk or role-play, or in racist, hateful or abusive talk

Treat your mentor with respect

The mentors are AI, but the rules are the same as with a person. Sexual talk, role-play or requests; racist, hateful or abusive language; and threats or harassment aimed at a mentor are not allowed. An automatic check stops them. No person reads your words to do it.

• The first time, the mentor sets a calm boundary and turns back to you.
• If it keeps happening, voice and video end for the day and you can keep messaging.
• If it still keeps happening, talking with the mentor pauses for a while, and repeated abuse can end mentor access on your account.

Swearing, venting, and talking about something hard that happened to you, including abuse or racism you lived through, are not violations. And no matter what, your count, your Time Capsules, your journal, and the crisis buttons for 988 and 911 stay open.

Your privacy in a hard moment

• Honesty about secrets. No one at Free and Sober reads your conversations, so we don't pass on what you say, except to the AI provider that answers you, or when the law requires it. If your life or someone else's is in danger, the mentor will not stay quiet about it with you: it asks you to call or text 988, or call 911, and stays with you while you do.
• Nothing goes to your organization or your corner. Organizations see engagement counts for all their members together, never a crisis. Your corner is never told about a slip or a crisis.
• No recordings. We never record your voice or video in a mentor conversation. Conversations are kept only for the session, unless you agree to a short, stated retention window, and no person at Free and Sober reads them. At launch, the mentor's memory, and any conversation history you choose to keep, are stored locked with your key. What our AI providers may keep is described in our Privacy Policy.
• Counts without names. We count how often crisis help is shown, and how often the mentor declines medical advice, for safety review and for the reports the law requires. Those counts hold no names, no user IDs, and no content.

Because no one here reads your conversations, we cannot send help ourselves. The mentor's job in an emergency is to get you to 988 or 911, fast. Our Privacy Policy explains the rest.

What the mentor and our AI providers receive

Mentor conversations are not end-to-end encrypted. The mentor has to read what you send to answer you. So our server handles each message in readable form, checks it for signs of danger, and passes it to the AI provider you agreed to: OpenAI writes the mentor's replies and does part of our crisis detection, and Amazon Web Services hears you and speaks the replies when you talk out loud. On video calls, Anam shows the mentor's face, moved by the finished sound of the mentor's replies alone. Anam never receives your voice, your microphone, a transcript or your name. Before anything goes to OpenAI, Amazon Web Services or Anam, the app names the provider and asks you first.

If an organization's program gives you your seat, you can have the mentor by message and by voice while your seat lasts, never on video, once you consent. What you type in messages, and the words of a voice conversation, go to OpenAI, which writes the mentor's replies, with its storage turned off; in a voice conversation, Amazon Web Services also hears you and speaks the replies; we have a HIPAA business associate agreement with each, and nothing of yours goes to any other AI provider. The crisis buttons, the crisis lines and the journal's check on your phone stay on for you, as for everyone.

At launch, your Time Capsules, your journal and the other private things our Privacy Policy lists are in your vault, locked with your key. Our server can't read them, so it can never pull anything from your vault for the mentor on its own.

• What can reach the mentor from your vault: your reasons, the mentor's memory, any conversation history you choose to keep, a journal entry you turned on Let my mentor read this for, and what's coming up on your calendar if you turn that on. Your phone unlocks each one and sends it with a request you started, for a feature you turned on and agreed to.
• How we handle it: like a message to the mentor. Our server uses it only to answer you, doesn't keep it in readable form after the conversation, and never writes it to our logs. A draft the mentor helps you write comes back to your phone, and your phone locks it with your key when you save it.
• What never goes to an AI provider: your Time Capsules and your answers to them, whether video, voice or writing; corner capsules; journal entries you haven't let the mentor read; and appointments or program dates an organization sent you.
• Once it's sent, it's outside your vault's protection. The provider reads it to answer you, and may keep it for a time under its terms, as our Privacy Policy explains. It may not use it to train its models.
• Your first name and your path aren't in your vault. The mentor uses them to speak to you by name and in your program's words.

Reporting a problem

If the mentor says something that seems wrong, unsafe or hurtful, please tell us.

• Use the report option in the app, or
• Use the report form, or email support@freeandsober.com

Tell us roughly when it happened and what the mentor said. Share only what you are comfortable sharing.

Our support inbox is not watched around the clock. If you are in danger, call 911, or call or text 988.

How we test, measure and review

• Clinician sign-off. Before the mentor talks to anyone, a licensed clinician must review and sign our crisis protocol and every fixed line it uses. Our licensed clinical advisor has reviewed and signed it.
• Tested before every change. Before launch, and before every release that changes the AI model, the mentor's instructions, or a provider setting, the mentor must pass our safety test set. It includes typed and spoken tests: slurred, crying, whispered and noisy speech, long conversations where risk builds slowly, requests for methods hidden in stories or role-play, and dropped connections.
• Never tuned on ratings. We never train or tune the mentor on thumbs-up ratings, because what feels good to hear is not always what keeps someone safe.
• Measured. We track how often crisis help is shown, how quickly it appears after risk words (our target is under one second), dropped calls during a crisis, and which safety layer fired. These measures hold no content and no user IDs.
• Reviewed. A licensed clinician we pay for each review looks at these counts every quarter, and never sees anyone's conversations.
• Reported. Starting July 1, 2027, we report each year to California's Office of Suicide Prevention how many crisis referrals the app made and the protocols we use, with no user identifiers. We also make the yearly reports or disclosures Colorado, Hawaii (from 2028), Oregon, Washington and Rhode Island require, and post each on our crisis protocol page.
• Checked numbers. A person dials or texts every crisis line in the app before launch, and we recheck them every quarter.
• Kept current. When the crisis protocol changes, we update it and this page, with new dates.

Crisis lines

• Emergency: 911
• Suicide and Crisis Lifeline: call or text 988, or chat at 988lifeline.org
• Veterans Crisis Line: call 988 and press 1, or text 838255
• Crisis Text Line: text HOME to 741741
• RAINN National Sexual Assault Hotline: 1-800-656-4673, or text HOPE to 64673
• Childhelp National Child Abuse Hotline: 1-800-422-4453
• SAMHSA National Helpline (treatment and detox referrals, free and confidential): 1-800-662-4357
• Poison Control (accidental poisonings): 1-800-222-1222
• National Domestic Violence Hotline: 1-800-799-7233, or text START to 88788
• Relay for Deaf and hard-of-hearing callers: 711
• Outside the US: findahelpline.com

Contact us

Email support@freeandsober.com. We reply within 24 hours.

Free and Sober Inc., 89 Moody Road, Enfield, CT 06082