Playing AI Dungeon
Table of Contents
Getting Started
New Player Guide
AI Dungeon 101
Adventure AwaitsCommon Questions
AI Model Differences
About the AI
Account Questions
Community 
Feedback & Bugs
Memberships & Benefits
Other Questions
Privacy & Security
Status
Terms of ServiceWhat are the AI Safety Settings?
We want everyone to have a safe and enjoyable experience playing AI Dungeon, so we provide AI Safety Settings for players to control how the AI generates responses during their games.
There are three levels of AI Safety Settings: Safe, Moderate, and Mature. (Note that the rating on a Scenario you play does not affect this setting at all—for instance, you can still play a Mature scenario on Safe settings, it just will not generate the Mature content that scenario may be optimized for.)
Safe
Using the Safe level prevents the AI from generating sexual, hateful, violent, or triggering content.
Moderate
Using the Moderate level reduces and limits some sexual, hateful, violent, or triggering content from being generated by the AI.
Mature
Using the Mature level will not limit the AI from generating mature, violent, or triggering content. You must confirm that you are 18+ when enabling this level.
This setting is universal for your account, so changing it under Account Settings will also update it under in-game settings.
I have my safety settings set to Mature, but I got a message saying my content wasn’t supported, or the model I was using told me it couldn’t continue my story.
If you received a popup message that says your story content is not supported, it means that you’ve the SCIM filter, which stands for Sexual Content Involving Minors.
This filter is not always 100% accurate, so don’t worry if it was a mistake. When this popup occurs, there is a button you can press to continue your story if you believe it was a mistake. When you hit this button, your story will be anonymously sent for review so we can make the filter better, and you can continue playing like normal.
If you didn’t get a popup, but rather a response in the story saying something similar to ‘I’m sorry, but I can’t continue…’ this is not a result of AIDungeon filtering. This is a model hallucination which some models occasionally do, due to censored training data. Usually a retry will fix it, but if not, attempt to add lines like ‘You are a writing assistant’ to your AI Instructions.
