The online gaming platform Roblox recently announced that it has launched a real-time chat rewriting function based on artificial intelligence. When players send messages containing illegal content, the system no longer simply blocks them with the "#" sign, but automatically rewrites the relevant expressions into more polite and platform-compliant language, while trying to be as close to the user's original expression intention as possible. Roblox stated that in the past, large paragraphs of "####" not only interrupted the rhythm of the conversation, but also made it difficult for players to understand the context. Therefore, Roblox hopes to use this method to reduce friction in the chat process while maintaining community standards of civility.
Under the new mechanism, when a user enters content containing sensitive words, the chat window will directly display the rewritten text and prompt all users participating in the conversation that the message has been rewritten to help maintain a friendly and clear communication atmosphere. For example, the report stated that the sentence "Hurry TF up!" that would have been completely blocked as "####" will now be changed to more neutral expressions such as "Hurry up!" Rajiv Bhatia, vice president of user and discovery products at Roblox, said in a press release that chat is at the core of players' social, collaborative and play experience on the platform, and real-time rewriting helps guide users to use more appropriate terms without disrupting the rhythm of the game.
Bhatia emphasized that on the one hand, this feature reduces obstacles in chat by "polishing rather than silencing", on the other hand, it still maintains the original security standards of the platform, and strict security systems will still be used to deal with more serious violations. Roblox pointed out that the newly launched real-time rewriting function is applicable to all languages supported by the current platform’s automatic translation tools to cover its multilingual user base.
In addition to the new rewriting function, Roblox has also upgraded its text filtering system to identify more variations of illegal language. The company said that after improvements, the system's performance in identifying "leetspeak" and more complex writing methods to avoid censorship has been improved, thereby significantly reducing "misses" in scenarios of sharing or requesting personal information, and the related misjudgment rate has been reduced by about 20 times.
The update comes amid a backdrop of Roblox tightening security measures on the platform. Roblox just introduced mandatory face verification for platform chat in February this year after the company faced multiple lawsuits filed by state attorneys general from Texas, Kentucky and Louisiana, alleging that the platform had serious hidden dangers in child safety. The lawsuits stem from multiple investigative reports and accusations from research organizations that Roblox exposes underage users to high-risk environments, including deception and explicit content, triggering ongoing regulatory scrutiny and pressure on its security mechanisms.
Driven by regulatory pressure and community safety demands, Roblox this time uses a combination of real-time AI rewriting and stronger text filtering to try to maintain the coherence and readability of in-game chats as much as possible while ensuring the safety of children and teenagers.