Report

Moderation

How moderation works

What is checked in chat, voice, images and video, when it is checked, and when a person looks.

Moderation on AI Angels has several layers: automatic checks on what you ask for, stricter checks before a photo or voice note is made, people who review ready-made content and handle reports, and your reports. No single check is perfect, which is why there is more than one.

The layers

1. Before she replies

Every message in a chat or group chat is checked against rules for the hardest limits: minors, family, non-consent and animals. In one-to-one chats, an AI classifier also reads each message, so a request in another language or in other words can be caught too.

2. Before a photo or a voice note

A request for a photo or a voice note gets a stricter check, which includes an AI review. If that check cannot run, the photo or voice note is not made.

3. On a call

What you say on a call is checked as you speak. A request that crosses a hard limit ends the call.

4. In the studio

When you create a character, an image or a video, your prompt is checked against a longer list of banned words and phrasings before anything is generated.

5. Ready-made content

Photos and voice clips prepared in advance for catalogue characters are approved by a person before a character can send them, unless that character's new photos have been set to go live without a one-by-one review.

6. Your reports

Anyone can report a character, an image or a user, with or without an account. A person reads every report.

What a person can see

Our staff can open accounts, chats and photo requests in our admin tools. We do this to handle reports and support requests, to look into safety concerns, to answer legal requests, and to check and improve how the characters reply. Other users never see your chats.

Automated systems read every message: to write her reply, to apply these rules, and to choose what she can offer, such as a photo. More on this is in Privacy at a glance.

When something is caught

  • In a chat, she says no in her own words and the conversation can carry on. She does not name the rule.
  • A refused photo or voice note is not made.
  • On a call, the call ends.
  • In the studio, you see a message that the prompt cannot be used.

Repeated attempts to get around the rules can lead to the account being suspended or closed. The usual reasons for a refusal, and what to change, are on Why was something blocked?

Self-harm and suicide

Messages in chats and group chats on the AI Angels website are also checked for signs that someone may harm themselves. What happens next is set out in the crisis protocol.

Taking content down

When a report or a check shows that a character, photo or video breaks the rules, we take it down. A catalogue character can be switched off and a photo taken off our pages. Reported illegal or non-consensual content is removed within 24 hours, as section 8 of the Terms says.

Last updated 18 November 2026. AI Angels is operated by AiAppDevelopment OÜ, Tornimäe tn 7, Kesklinna linnaosa, 10145 Tallinn, Estonia. [email protected]

Folds the footer down to its legal links. Turn it off to see every link.