Roblox now employs AI moderation to eliminate harmful content before it can reach users.
Roblox is enhancing its safety measures with a new AI moderation system.
If you’ve played Roblox, you’re likely aware of the platform’s chaotic and unpredictable nature. In response, Roblox is leveraging this unpredictability to revamp its content moderation approach.
The platform has introduced a real-time multimodal AI moderation system that analyses entire in-game scenes instead of just individual occurrences to identify content that may have evaded previous checks.
How does Roblox’s new AI moderation function?
In contrast to older moderation tools that assess one object at a time, the new multimodal system evaluates entire scenes, incorporating avatars, text, and 3D objects, to ascertain whether the overall combination violates Roblox’s Community Standards.
For example, if someone uses a free-drawing tool to create an offensive symbol—something that wouldn’t trigger a single-item check but is clearly problematic within context—this system is designed to identify it.
When repeated violations are identified in a single game instance, the system disables just that server rather than the whole game. Since its implementation, about 5,000 servers have been taken down daily.
Roblox aims to monitor all playtime with this system. In addition to server suspensions, it is also working on tools to pinpoint and remove individual offenders without negatively affecting the experience for others.
What does this entail for Roblox creators?
Creators won’t be left uninformed. A new chart on the existing Creator Dashboard displays the number of your game servers taken down each day due to misconduct.
Notable increases in this number can indicate issues worth looking into, providing creators an opportunity to review and modify aspects like custom emotes or in-game building tools before the situation escalates.
A new certification initiative for improving online gaming
Roblox is also addressing a broader industry concern. In collaboration with Keyword Studios and Riot Games, it is co-developing the DLC Leadership Program for online community managers and moderators.
Research psychologist Rachel Kowert, who serves as Research Director at Games for Change, is overseeing the academic component. The aim is to establish standardized, evidence-based training for individuals managing gaming communities, which the industry has previously lacked.
Roblox has been strengthening its safety measures for some time, including parental controls and age-based chat filters, with this latest update being its most ambitious effort to date.
Other articles
Roblox now employs AI moderation to eliminate harmful content before it can reach users.
Roblox's updated AI moderation system examines entire game scenes in real time, identifying harmful content that many previous systems would have overlooked, and is currently shutting down 5,000 servers each day.
