Roblox Deploys AI to Filter Banned Chat in Real-Time
▼ Summary
– Roblox is launching an AI-powered chat feature that rephrases banned words into respectful language in real-time, replacing the old system that used hash symbols.
– The new system aims to keep conversations clear and on track by preserving the user’s original intent, such as changing “Hurry TF up!” to “Hurry up!”.
– A company executive states the feature helps maintain community standards while reducing friction in chat, which is central to connection and gameplay on the platform.
– Roblox is also improving its text filter to better detect evasive language like leetspeak, reducing false negatives for sharing personal information by 20 times.
– This update follows recent safety measures, including mandatory facial verification for chat access, implemented after lawsuits from several states over child safety risks.
Roblox is rolling out a new artificial intelligence system designed to instantly rephrase inappropriate chat messages into more civil language, moving beyond its previous method of simply blocking words with symbols. This real-time feature aims to maintain the flow of conversation while enforcing community standards and guiding users toward respectful communication. By interpreting the intent behind a message, the AI seeks to reduce the disruptive strings of hashtags that previously made chats difficult to follow.
The platform’s older text filter would replace banned phrases with the “#” symbol. While effective at censorship, this often resulted in messages becoming a confusing series of symbols, breaking the natural rhythm of discussion. The upgraded system now analyzes the context and attempts to rephrase the message into something acceptable that retains the user’s original meaning. For instance, a phrase like “Hurry TF up!” will no longer appear as “####” but will be automatically changed to “Hurry up!” A small notification informs all participants in the chat that a message has been altered to keep the discussion appropriate.
Chat is a fundamental component of how people connect and collaborate within the Roblox universe, according to company executives. The introduction of real-time rephrasing is intended to keep gameplay and social interactions seamless while subtly steering language toward community guidelines. This method is seen as a way to decrease friction during conversations without compromising on safety. The underlying safety systems continue to monitor for and act upon more severe violations, ensuring that rephrasing is just one layer of a broader protective framework.
This new capability is integrated with Roblox’s existing automatic translation tools, making it functional across all supported languages on the platform. Alongside the rephrasing feature, the company has also enhanced its core text-filtering technology. Early data indicates the system has become significantly more adept at identifying creative attempts to circumvent filters, such as the use of leetspeak or character substitutions. These improvements have reportedly led to a dramatic reduction in false negatives, particularly in cases involving the sharing or solicitation of personal information.
The announcement follows other recent safety initiatives from Roblox, including the implementation of mandatory age verification for users accessing certain social features. These measures come amid increased scrutiny from regulators and legal challenges concerning child safety on digital platforms. The company emphasizes that these continuous upgrades to its moderation tools are part of a committed effort to foster a secure and positive environment for its global community.
(Source: TechCrunch)
