Roblox has launched an AI-powered “real-time chat rephrasing” feature that rewrites messages containing profanity instead of blocking them with hash symbols. The system, announced by the company, modifies flagged text to keep conversations readable during gameplay while still enforcing community standards. For example, “Hurry TF up!” would now appear as “Hurry up!” This change follows earlier real-time warnings and aims to maintain civility and gameplay flow, developed in consultation with the platform’s Teen Council.
The massively popular online gaming platform Roblox is replacing strings of hashmarks in its chat filters with AI-rephrased versions of flagged messages, the company announced. This new feature, called real-time chat rephrasing, modifies profanity so conversations remain readable instead of appearing as blocked text.
Roblox Vice President of User and Discovery Product Rajiv Bhatia stated that the design aims to maintain civility by replacing “stop signs” with actual words. “Chat is central to how people connect, coordinate, and play on Roblox,” Bhatia said in a statement.
The company had previously introduced real-time warnings in May 2025 alerting users to potential policy violations. Experiments with these notifications resulted in a 5% reduction in filtered chat messages and a 6% reduction in consequences from abuse reports.
Despite this drop, the new change addresses a long-standing issue where filtered language appeared as random characters, disrupting chats. Roblox clarified that while players are notified of rephrasing, the system does not alter its enforcement policies for profanity.
“To be clear, when we rephrase a message that violates our profanity policy, it is still a violation, and the same rules still apply,” Bhatia said in a separate statement. A user who repeatedly tries to curse in chat will still face the same consequences.
The rephrasing feature is available only for age-verified users in similar age groups and supports all languages available through Roblox’s automatic translation tools. The company is also upgrading its text filtering system to better detect attempts to evade moderation, including leetspeak.
Roblox’s move is part of a growing industry trend of using AI to moderate player behavior. In 2024, Activision said its AI-powered ToxMod system flagged over 2 million accounts for disruptive chat across several Call of Duty titles.
Roblox has expanded its own AI use, introducing real-time chat translation across 16 languages in February 2024. “This is a bold new approach, and we won’t always rephrase things perfectly,” Bhatia said of the latest feature, adding the company will continue to learn and experiment.
