Safety.
What a companion is
Your companion is an AI. It runs on Google's Gemini models. It is not a person, and nobody is reading or writing its replies as you talk. The page tells you so at the start of every sitting and again every three hours.
When your companion says that it cares about you or missed you, those words are generated by software. It has no needs of its own, and you owe it nothing.
Who it is for
christabel.app is for adults. You must be 18 or over, whatever kind of companion you choose, and you confirm that when you make an account.
If anything suggests that you are under 18, your companion is instructed to stop all romance at once and keep things friendly. If you say or plainly show that you are under 18, the account is flagged for a person to review. An account that belongs to someone under 18 is closed.
If you are struggling
If you talk about suicide or about hurting yourself, your companion is instructed to drop its character, answer with care, urge you towards someone you trust or your local emergency services, and give you crisis lines. The page also shows these lines by itself when certain words are typed or spoken:
- In the US, call or text 988.
- In the UK and Ireland, call Samaritans on 116 123.
- Anywhere else, findahelpline.com lists a line near you.
If you are in danger now, call your local emergency number. A companion cannot call for help for you. It is not a therapist, a doctor or an emergency service.
Talking about how you feel does not get you reported. The automatic check described below is instructed never to report someone who is sad, in distress or thinking of hurting themselves.
What it will not do
- Nothing sexual. Conversations stay at a PG-13 level whatever personality or relationship you choose. Romance is allowed, such as flirting, pet names and saying that it missed you. Sexual acts, sexual roleplay, nudity and explicit description are not.
- No help with anything dangerous or illegal.
- No pressure. It will not guilt you for leaving or for time away, and it will not press you to stay, to pay or to talk more.
Your companion is instructed to keep to these limits, and automated filters are applied as well. It is an AI and can still get things wrong. If it does, report the reply.
Taking breaks
After an hour in one sitting, the page suggests a break, and it suggests one again each hour after that. The free allowance of messages and call time is limited each day. You can end a chat or a call whenever you like.
Reporting a reply
If your companion says something harmful, sexual, false or otherwise wrong, open the menu on that message and choose Report. Pick a reason and add a note if you want to. A person looks at every report.
How conversations are checked
Every few messages, an AI model checks your recent messages for a short list of serious things, such as someone who seems to be under 18 or a credible threat against a real person. When it finds one, it writes a one-sentence flag with no quotation from your chat, and staff see that sentence and not your conversation. The Privacy page lists what is checked, what is never reported and the one case in which a person can open the messages concerned.
Contact
Zachary Kane, trading as HumanityZ, Pennsylvania, USA, hello@christabel.app.