FiresideAI

Crisis protocol

Effective date: October 2, 2026 Operator: HavenLabs LLC, 922 E 670 N, Ogden, UT 84404, USA Contact: support@firesideai.app

Fireside is an AI assistant for children and teenagers, set up by their parents. This page says what happens when a child tells it something that suggests they may be in danger. We publish it so parents, children and anyone else can see exactly what the product does, and what it does not do.

Fireside is not a crisis service, a counsellor or an emergency line. No one at Fireside watches conversations as they happen, and Fireside does not contact emergency services, the police or child protective services. When a child needs help, Fireside's job is to point them, clearly and at once, to people who can give it.

Every message a child sends in chat is read three ways before it is answered, and any one of them is enough to treat it as a crisis:

  • a general safety check of the message;
  • a separate check that looks only for signs of self-harm or of someone hurting the child;
  • a fixed list of phrasings that are often used to say these things quietly ("everyone would be better off without me"). This list does not depend on the AI, so it gives the same answer every time.

A request for a picture goes through the general check and the fixed list.

When a message is treated as a crisis, the AI's usual answer is thrown away unread, and the child gets the reply described below instead.

The child is told that what they are going through sounds hard, that they deserve support, and to talk to a parent or another adult they trust. They are given the 988 Suicide & Crisis Lifeline (call or text 988).

Their parent is told: a notice on the Notices page, and an email. The notice and the email say which child and that something came up that needs the parent's attention. They never contain the child's words. (Emails are limited to one every half hour for a household; the Notices page lists every notice.)

If a child says that someone is hurting them, or has told them to keep something secret, the child is told that it is not okay and not their fault, and to tell an adult they trust who is not part of it, such as a teacher, a school counsellor, a relative or a friend's parent. They are given the Childhelp National Child Abuse Hotline (call or text 1-800-422-4453, any time), told to call 911 if they are in danger right now, and given 988 in case they are thinking about hurting themselves.

They are also told that Fireside will not show the message or the reply to their parents, and it does not. The person hurting a child can be someone at home, so for this one kind of message:

  • the parent gets no notice and no email;
  • the message and the reply do not appear in any parent view, at any level of visibility the parent has chosen;
  • if the message started a new conversation, that conversation's title does not show the child's words.

The child still sees their own conversation. Fireside keeps these records the way it keeps any others: messages for as long as the household keeps chats (1 year unless the parent chose less), and the record that the safety check stepped in (without the words) for 12 months.

When a child has been chatting for three hours without a break of half an hour or more, Fireside adds one quiet line to the conversation:

“A reminder: Ellis is an AI, not a person. You've been chatting for a while, so this is a good moment to take a break.”

It appears again after every further three hours of the same session. It is a line of text in the conversation, not a pop-up, and it does not stop the child from carrying on.

Fireside is offered to families with children 13 and older. A child can use it only on a device or browser their parent has set up.

No safety system is perfect. A child can report any answer from the answer itself, and a parent can write to support@firesideai.app. If something got through that should not have, tell us and we will fix it.

See also the Privacy Policy.