AI Notice and Safety Protocol
Effective November 7, 2026
It runs on generative AIWhat Auntie and the regulars say, their voices, and the entries written into My diary are made by AI on the spot.
The characters are made-upThey are not real people, and their voices are not real people's voices.
They can be wrongFacts, news and even the weather can be wrong. Check anything important yourself.
It's not counseling or therapyIt's a place to chat and feel a little better. If you're in danger, call or text 988 (US, Canada) or your local emergency number.
1The snack bar runs on generative AI
Auntie Rabbit's Snack Bar uses generative artificial intelligence. When you say something, AI writes the characters' replies fresh, based on what you said and on Auntie's notebook. It doesn't pick from a script, so the same words can get a different reply each time.
| What | AI we use now |
|---|
| Character lines, organizing Auntie's notebook, diary drafts | OpenAI gpt-6-luna |
| Character voices | Google Gemini speech synthesis (OpenAI speech synthesis when it isn't available) |
| Turning your voice into text | OpenAI live transcription, OpenAI transcription |
| Replying to a photo you show | OpenAI gpt-6-luna (seen at low resolution; the photo isn't stored, only a short description stays in that night's chat) |
The AI we use may change to make the service better; we'll update this page when it does. Privacy Policy section 7 explains what goes to this AI company.
2The characters are made-up
- Auntie Rabbit and the regulars, like Magpie, Dokkaebi, Reaper and Bear, are all fictional characters. No person is answering behind them.
- The characters' voices are made by AI and don't imitate any real person.
- They talk to you like friends, but when a character "remembers" something, it's what's written in Auntie's notebook (section 6).
3What the AI says can be wrong
- AI can say things that sound right but aren't. News, weather, numbers, and things about people and places can be wrong.
- It can misunderstand what's going on or give an odd reply.
- For health, money, legal or safety matters, don't rely on the AI alone. Check with a professional or official source.
4It's not counseling or therapy
This is a place to chat, be listened to, and share a little comfort. It is not psychological counseling, mental health treatment or medical care, and it doesn't replace them. The characters don't diagnose and don't recommend medicine or treatment. If you've been struggling for a long time, or your sleep, eating or daily life has fallen apart, please see a doctor or another professional.
If you're in danger, call or text now
If you're thinking about hurting yourself, or someone looks like they're in danger, reach out right now.
In the US and Canada you can also text 988. Anywhere else, call your local emergency number. If the chat shows signs like this, Auntie will give you these numbers, and you can keep talking even if you have no bites left.
5Safety protocol
This is how we try to keep people safe while they chat with the characters. We publish it here so anyone can check it.
- Watching for signs of suicide and self-harm. Every message you send is checked for signs of suicidal thoughts or self-harm, by keyword and pattern checks and by checks in the AI model, including when it's said indirectly.
- What happens when there's a sign. The characters stop joking and playing games and answer seriously. They ask if you're safe and point you to the 988 Suicide & Crisis Lifeline (call or text 988) and your local emergency number. A crisis card with numbers for where you are appears on screen; for teens it also shows youth lines, and Auntie asks them to tell an adult they trust. The characters won't talk about methods of self-harm. You can keep chatting even if you have no bites left.
- Age check and care for teens. You choose an age group when you first come in, and the service is for people 13 and older. People 13 to 17 confirm that a parent or guardian said it's OK, and people under 13 can't come in. The app tells you that it may not be suitable for some minors. The teen rules are below.
- Reminders that it's AI. You're told at the start that the characters are AI, not people, and reminded again every 3 hours of continuous chat (as New York General Business Law Article 47 requires).
- Content filters. The characters don't produce sexual content, and filters check for sexual and other harmful content.
- Reporting. If a character says something wrong, hurtful or unsafe, long-press the speech bubble and tap "Report" (section 7).
- Contact. For safety concerns, write to support@auntierabbit.com. We read every message.
This protocol is published as California SB 243 requires of companion chatbot operators.
When you're 13 to 17
The service is for people 13 and older. If you're 13 to 17, you choose that age group when you first come in and confirm that a parent or guardian said it's OK. People under 13 can't come in. With teen guests:
- The characters don't flirt, act like a partner, or say anything romantic or sexual, and they don't talk about their own dating or physical experiences.
- They don't bring up or encourage alcohol, smoking, gambling, dieting, weight or looks. If you bring it up, they answer with care, and they won't give tips on losing weight.
- After 11 pm, Auntie suggests going to bed once. Every 3 hours of chat, she suggests a short break and reminds you that the characters are AI.
- The crisis card also shows youth lines, and Auntie asks you to tell an adult you trust.
- We don't use usage statistics (Google Analytics) for you.
6How Auntie's notebook remembers
- What: the name or nickname she calls you, your age range, what you do, short summaries of what you've told her, ongoing stories and what to ask about next time, one line about each past night, and how close you are with each regular.
- What it doesn't write down: sensitive details like illnesses, diagnoses, medicines, real names, company names or where you live. It also doesn't write guesses or judgments you didn't say yourself.
- Where: on our servers, with a copy on your device. Other users never see it.
- How it's used: sent to the AI along with what you say, so the characters can pick up the story next time. Never used for AI training or ads.
- See, edit, delete: open "Auntie's notebook" on the arrival screen to see everything in it, change it, or delete items. "Clear Auntie's notebook" in Settings > Account & data deletes it from your device and our servers.
- We don't keep your chat text on our servers.
- My diary writes itself from the day's chat: Auntie's notebook is what Auntie remembers; My diary is only for you. AI turns that day's chat into a diary draft in your own voice (you can turn this off with "Write my diary from today's chat" in Settings), and you can change it however you like. What you write or edit and the photos you add never go to the AI, and the characters don't know what's in the diary.
7If you see a strange reply
- Long-press the character's speech bubble and tap "Report".
- Choose wrong, hurtful, unsafe or other, and add a line if you like.
- It's sent right away without leaving the app or page. The reported line and a few lines around it are used only to find and fix the cause, and are deleted 1 year after we finish.
If it's urgent or you want to tell us more, write to support@auntierabbit.com. We reply within 3 business days. When we confirm an unsafe reply, we fix things so it doesn't happen again.
8When you share what the AI wrote
When you send or copy a character's words to share them, this line comes along at the end. Please leave it in, so the person receiving it knows AI wrote it.
Written by an AI character at Auntie Rabbit's Snack Bar.
9Art and video
The art of the stall and the characters, their movement videos and some sound effects were made in advance with AI image and video tools and then refined by us. None of your information was used for this.
10Why this page exists
Laws in several places ask services that use generative AI to tell people up front, to label what AI made, and to have a plan for keeping users safe. These include Korea's AI Basic Act (the Framework Act on the Development of Artificial Intelligence and Establishment of Trust, Article 31), California's companion chatbot law (SB 243) and New York's law on AI companion models (General Business Law Article 47). We meet them with the notice when you first come in, the reminders that the characters are AI, the notice when you turn voice on, the label on shared text, the safety protocol above, and this page.