Roblox is deploying artificial intelligence to modify inappropriate chat messages in real time and upgrading its moderation tools to better identify abusive language and attempts to bypass platform rules.
Gaming platform Roblox has introduced a new artificial intelligence-powered feature aimed at improving communication among players on its platform. The update, called real-time chat rephrasing, is designed to make conversations during gameplay smoother while ensuring interactions remain respectful.
The feature automatically modifies inappropriate language in chat messages instead of simply blocking them. Previously, when the platform’s filters detected profanity or policy violations, the message appeared as a series of hash marks, which often disrupted conversations and made chats difficult to follow.
With the new system, the AI rewrites such messages into more suitable language while keeping the original meaning as close as possible. For instance, a message containing offensive wording may now appear in a toned-down version rather than being completely hidden.
AI helps maintain flow of conversation
The company says the new approach aims to reduce interruptions in chat and allow players to communicate more naturally while still maintaining community standards.
According to Rajiv Bhatia, chat plays a vital role in how players interact, coordinate gameplay, and build connections on the platform. The AI-powered rephrasing feature is designed to guide users toward more appropriate language while ensuring discussions continue smoothly.
Whenever a message is modified, participants in the conversation are notified that the text has been adjusted to maintain civility. The feature is currently available for in-experience chats between users whose ages have been verified and who are in similar age groups. It also supports all languages currently offered through Roblox’s automatic translation tools.
Stronger filters to detect harmful content
Alongside the rephrasing feature, Roblox has also upgraded its text filtering technology to better detect attempts to bypass its community guidelines. The improved system can identify variations such as “leet-speak,” where letters are replaced with numbers or symbols to avoid detection.
Early testing suggests the enhanced filters have significantly improved the detection of attempts to share personal information, such as social media handles or phone numbers.
Roblox says these improvements represent another step in its broader effort to create a safer and more positive online environment while allowing players to communicate freely during gameplay.
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
