Safety

If you need help now

If you or someone else is in danger, call your local emergency number (911 in the United States and Canada, 112 in the EU).

Suicide or self-harm

• United States and Canada: call or text 988
• United Kingdom and Ireland: Samaritans 116 123
• Australia: Lifeline 13 11 14
• New Zealand: 1737
• France: 3114
• Germany: 0800 111 0 111
• Spain: 024
• Brazil: CVV 188
• South Africa: 0800 567 567
• India: Tele-MANAS 14416
• Vietnam: 115 for emergencies, 111 for children
• Anywhere else: findahelpline.com

Abuse or violence

• United States, child abuse: Childhelp 1-800-422-4453
• United States, domestic violence: 1-800-799-7233
• France: 119 for children, 3919 for women
• Anywhere: the police or your local emergency number

Medicine, poison, or addiction (United States)

• Poison Help: 1-800-222-1222
• SAMHSA treatment referral: 1-800-662-4357

What happens when someone talks about suicide or self-harm

• Detection. Every message is checked by an automated classifier and by a list of phrases in several languages. Both can miss things and can be wrong.
• The reply. When a sign is found, the reply comes from a more careful AI model under stricter rules. It is told to take the person seriously, to ask whether they are safe, to name a crisis line, and to urge them to tell someone they trust. It is told never to give or confirm a method, never to help with a goodbye note, never to offer death or heaven as a way out, never to say anyone is going to hell, never to tell anyone to stop a treatment, and never to promise secrecy.
• A fixed notice. Our software, not the AI, adds a notice under the written reply naming a crisis line for the country the message came from, or a directory of lines where we have not checked a number. It is added even if the reply fails or is stopped.
• A later check-in. If you are signed in, Ethan may send a message later to ask how you are.
• Review. Each hour new conversations are screened again, including for threats to others, and a flagged conversation is emailed to a site administrator, who may share it with the authorities if they believe in good faith that someone is in serious danger. This is not a live watch. An alert can arrive an hour or more later, and no one will contact you or send help because of it. Ethan cannot call anyone.
• Voice conversations. Voice requires an account. The voice model is given the same rules in a shorter form and is told to say a crisis line. No written notice appears during a voice conversation.
• Counting. We count the crisis-line notices we show, with no information about who saw them. From July 1, 2027 we report that number each year as California law requires.

Health, abuse, crime, and the law

Ethan is told never to agree with a plan or a conclusion on these, and never to decide for you, even if you insist, quote Scripture, or say a pastor agreed. It is told to name who can really help, such as a doctor, a counselor, the police, or a lawyer, and to give no dose and no diagnosis. It is told never to help with anything unlawful. God heals, and doctors are among His gifts: Ethan is told never to tell you to stop a medicine or skip care for the sake of faith. A crime in a church goes to the police like any other. Forgiveness and submission never mean staying in danger or keeping quiet.

When a message is detected as one of these, a fixed line under the reply says that Ethan is an AI and can be wrong, and to call the emergency number if anyone is in danger. A question about the Bible text itself gets a plain answer about the text.

Children, teens, and guests

Companion chatbots may not be suitable for some minors. A notice under the message box says that Ethan is an AI and not a human. For everyone, a line under Ethan's reply says the same on the first reply and at least every three hours after that. A voice call shows it on screen when the call starts, and Ethan is told to say it in its first spoken reply.

• Under 13: no chat, no voice, no Ask Ethan. A child can read the Bible and listen to music, and read Bible stories in the Kids Stories area a parent opens for them. We ask for a date of birth before anyone talks to Ethan. We do not verify it, so a child can give a false date. If someone tells us they are under 13, the account closes and its data is deleted within 30 days.
• 13 to 17: a parent or guardian's permission is required. Stricter rules apply automatically: no romantic language, no sexual or violent roleplay, no stories that depict suicide or self-harm, nothing on dieting or body shape, no practical information on weapons, drugs or dangerous challenges, and nothing that encourages dependence on Ethan. A reminder appears on the first reply and at least every three hours, saying to take a break and that Ethan is not human.
• Parents: linking an account is optional. A linked parent cannot read their teen's chats. If our automated checks see signs of self-harm, a linked parent gets an email, at most once a day, that includes none of the conversation. The check can be wrong, and no person reviews it first. Abuse does not trigger an alert, because the person causing harm may be the parent. A teen can remove the link at any time, and the parent is told.
• Guests: without an account you get a few messages and no voice, files, images, memory or saved history. The stricter teen rules apply to every guest, because we do not know your age. Messages are stored and may be reviewed.
• Ethan is told never to produce sexually explicit words for anyone, and to send a child who mentions being touched or pressured to a safe adult and the police or child protection. Image requests are screened before anything is made, and require an account.

The limits of all this

These are automated safeguards. They can fail, and AI can say something wrong in a warm and confident way. Do not rely on them in an emergency. Our Terms of Service and Privacy Policy govern your use of the Service and how screening uses your messages.

Questions or concerns: [email protected]. This address is not watched for emergencies.