Roblox launches real-time AI chat rephrasing to filter out banned language

Summary

Roblox has introduced a real-time AI-powered chat rephrasing feature that automatically replaces banned words with more respectful language, moving beyond its previous approach of simply masking inappropriate content with hash marks. The new system uses AI to rephrase messages to maintain their original intent while filtering out offensive language — for example, converting "Hurry TF up!" to "Hurry up!" while notifying all chat participants that the message has been modified.

According to Rajiv Bhatia, VP of User and Discovery Product at Roblox, the feature aims to reduce friction in conversations while maintaining community standards, as the previous hash-mark system often made chats hard to follow and disrupted gameplay. The AI rephrasing works across all languages supported by Roblox's automatic translation tools, though the company's existing safety system remains in place for more serious violations.

The announcement comes alongside upgrades to Roblox's text-filtering system, which has improved detection of leetspeak and sophisticated filter-bypass attempts, reducing false negatives for personal information sharing by 20x. This launch follows recent mandatory facial verification requirements introduced after multiple state attorneys general filed lawsuits against Roblox over child safety concerns, including allegations of grooming and exposure to explicit content on the platform.

• The feature launch follows implementation of mandatory facial verification and multiple child safety lawsuits filed by state attorneys general

Editorial Opinion

Roblox's AI rephrasing represents a more sophisticated approach to content moderation that balances safety with user experience, addressing a longstanding complaint about hash-mark filtering making conversations incomprehensible. However, the timing — immediately following child safety lawsuits — raises questions about whether this technological solution adequately addresses the deeper structural safety issues on the platform. While reducing friction in casual conversations is valuable, the real test will be whether this AI system can effectively prevent the grooming and exploitation that prompted regulatory scrutiny, or if it's primarily a UX improvement being positioned as a safety enhancement.