Managing content safety in Questsmith
Other
As part of our major platform update, we overhauled our AI Safety Settings. Previously, many players felt older safety controls did not provide the level of control they expected, and some still encountered uncomfortable AI outputs.
After reviewing this feedback, we recognized that content safety is a difficult challenge and improved our approach with a specialized technology partner, HiveAI, to provide better safety scoring and control.
When players interact with AI, model outputs are sent for safety scoring. If an output does not meet the player-selected threshold, we provide an alternative generation or notify the player that there was an issue. This is also applied to Scenario Images and selected Scenario Content.
To protect privacy and security, API requests are encrypted, processed, and deleted quickly. Data sent for safety scoring is not stored, reviewed, or used to train partner models. It is only used to generate a gameplay safety score.
Our AI Safety Setting has three levels: Safe, Moderate, and Mature. These map roughly to familiar movie ratings: Safe ~ PG, Moderate ~ PG-13, and Mature ~ R/18+.
We also improved Discover and Search safety controls using content ratings: Everyone, Teen, Mature, and Unrated (explicit content).
The AI Safety Setting and Search Filter operate independently because some players prefer creating NSFW content privately while avoiding NSFW content in browsing.
Defaults are Safe for AI Safety and Teen for Search Filter, and players can change both settings anytime.
These controls are designed to give players more personalization and safety options while keeping flexibility for different play styles.