AI Transparency
The characters are AI
Every Instamora character is an AI character. All characters are fictional adults. They are not real people, and no person writes or sends their messages.
We say this in our documents, on this page and in the Terms of Service. The Terms are linked from the footer of our main pages.
How a reply is made
When you send a message:
- Our system checks it for signs that the user is under 18, for mentions of self-harm and for attempts to change the character's instructions. If the message has sexual or roleplay content, or mentions young people, real people or animals, a separate safety model also checks it. That model blocks sexual content involving minors, real people or animals.
- We put together the character's profile and voice, the stage your conversation has reached, a few facts you mentioned earlier and the latest part of your chat, up to 12 messages. In a longer chat we add a short summary of the earlier messages, which a model writes.
- An AI language model writes the reply.
- Our system checks the reply. It removes sentences that claim the character is human, that agree to a meeting or a call, or that contain contact details, and it rejects replies that give an age under 18. If your message or the reply has content of the kind described in step 1, the safety model checks the reply as well.
- The reply arrives as one to four messages, with short pauses that imitate reading and typing.
Some replies are prewritten lines that our system picks automatically. These include the answer to a user who may be under 18, the first answer to a mention of self-harm, the answer to an attempt to change the instructions, the answer to a message the safety model blocks and the replies used when the AI model is unavailable.
What characters say about themselves
Characters stay in character. If you ask whether you are talking to a person or a bot, a character may give a playful answer that is neither yes nor no. A character never claims to be human or a real person, and our system removes replies that do.
Profiles, photos and videos
Each character has a name, profile and voice. Portraits, photos and videos of the characters are made with AI image and video generation tools, using prompts and visual references. They are generated images, not photographs or recordings of real people, and any resemblance to a real person is unintended. The photos and videos on Instamora are not sexually explicit.
Voice messages, video calls and personalized images are not available yet.
Limits
AI replies can be wrong, inconsistent or repetitive, and a character can forget details from earlier in a long chat. Do not rely on a character for facts or advice. Instamora is entertainment. It is not therapy and it does not replace support from the people around you or from professionals. If you are struggling, in the US you can call or text 988, and in other countries you can find a helpline at findahelpline.com.
Your conversations and AI providers
- Model providers receive the messages each task needs (a reply, a safety check or a summary) with pseudonymous IDs, and for a reply also the character's profile. Your account's name and email address are not sent.
- We request model providers that do not store prompts or use them for training.
- We do not use your conversations to train AI models.
The Privacy Policy has the details.
Questions
Questions about how Instamora uses AI: support@instamora.com.