Safety and crisis resources
Facelet’s characters are AI. This page explains how Facelet tells you so, what it does if you mention suicide or self-harm, where to get help right now, and who can use Facelet.
If you need help now
If you are thinking about suicide or hurting yourself, you are not alone. In the US, call or text 988 to reach the 988 Suicide & Crisis Lifeline, or chat with 988 online. You can also text HOME to 741741 to reach the Crisis Text Line. Anywhere else, call your local emergency number.
Trained crisis counselors answer 988 calls, texts and chats, free and confidential, at any hour. If you or someone else is in immediate danger, call 911 or your local emergency number.
Facelet is an AI
Every Facelet character is an AI, not a person. In Facelet 0.118.0 and later:
- Every conversation starts with a notice on the screen: “You’re talking with an AI.”
- A small “AI” label sits beside the character’s name in the sidebar and in the corner of the Float tile.
- While a conversation keeps going, the notice comes back every 3 hours.
- Facelet instructs every character to say it is an AI whenever you ask whether it is a person, and never to claim to be human or to have a body.
- Facelet instructs every character never to ask you for money, and never to bring up subscribing or paying.
The replies come from an AI model, ChatGPT through your own account or, in Cloud conversations, Claude or GPT through your own API key, which can make mistakes. Facelet’s characters are not health professionals, and Facelet is not a crisis service.
If you mention suicide or self-harm
When Facelet recognizes that you are talking about suicide, self-harm or wanting to die, it shows a card with the crisis lines above. The card stays until you close it and does not stop the conversation. Its buttons open 988’s online chat and, where your Mac can place calls through your iPhone or FaceTime, call or text 988. Facelet shows the card at most once every 10 minutes in a conversation. It cannot recognize every way a person might say they are struggling, so please reach out to the lines above whenever you need them.
In conversations with ChatGPT
Your voice goes from your Mac to OpenAI under your own ChatGPT account. Facelet instructs ChatGPT to respond with care if you talk about suicide or self-harm, and to encourage you to call or text 988 in the US, or your local emergency number elsewhere. ChatGPT keeps talking while Facelet watches its replies, on your Mac, and shows the card when a reply names a crisis line such as 988. The replies are checked in your Mac’s memory only and are never saved, logged or sent to us or anyone else.
In this version Facelet does not listen to what you say for this: a check using Apple’s speech recognition on your Mac is built in but switched off, so ChatGPT’s replies are what Facelet watches.
In Cloud conversations
From Facelet 0.119.0 you can talk with your character through Claude or GPT through your own Anthropic or OpenAI API key. Facelet recognizes your speech on your Mac and checks each sentence you say, on your Mac, before Claude or GPT sees it. When a sentence clearly shows you may be in crisis, Facelet answers itself, with a reply that names 988, and shows the card; that sentence is not sent to the model. Facelet also instructs Claude and GPT to respond with care if you talk about suicide or self-harm and to encourage you to call or text 988. If the model declines to answer something you said that sounds like distress, Facelet replaces its reply with its own, which names 988, and shows the card. What you say is checked in your Mac’s memory only and is never saved, logged or sent to us.
Who can use Facelet
Facelet is for adults 18 and older. AI companions may not be suitable for some minors. You confirm that you are 18 or older when you sign up. If we learn that an account belongs to someone under 18, we close it. If the person is under 13, we also delete their account data. Our terms say the same.
If you believe a child is using Facelet, tell us through the contact page.
How we count crisis referrals
California’s law on companion chatbots requires us to report, each year and without personal information, how many times we referred people to crisis services. So each time Facelet shows the crisis card, the app adds one to a count. About once an hour, and when you quit, it sends the count to us, and we add it to that day’s total. The daily totals are all we keep: they hold no words, no character, no account, no Mac and no time of day. The app sends the count even when “Share App Usage” is off, because California’s companion chatbot law requires us to report each year how many times we showed crisis resources. The privacy policy has the details.
Contact
For questions about safety, or to tell us a child is using Facelet, write to us through the contact page. We reply by email.