Roblox rolls out real-time AI chat rephrasing to keep gameplay conversations civil

Roblox is rolling out a new AI-powered feature that automatically rephrases profanity in chat rather than replacing it with strings of hashtags, the company announced today. The goal is to keep gameplay conversations flowing naturally while still holding the line on its Community Standards.

If you’ve ever played a game on Roblox, you know how disruptive it can be when chat fills up with “####” every time someone gets a little heated. The new rephrasing system works by detecting messages that violate its profanity policies and then translating those messages into cleaner language that still captures the user’s original intent. The classic example the company gave is “Hurry TF up!” becoming simply “Hurry up!” Everyone in the chat is notified that the message was rephrased, so there’s no mystery about what happened.

The feature is live exclusively for in-experience chat between age-checked users in similar age groups, and it works across all languages currently supported by Roblox’s automatic translation tools. It is not a blanket replacement for Roblox’s wider safety systems. For more serious violations, the multilayered enforcement structure the platform already has in place remains fully active.

Rajiv Bhatia, Vice President of User and Discovery Product at Roblox, said real-time rephrasing is designed to “reduce friction in chat while maintaining the standards that help keep our community civil.” The company consulted its Teen Council while building the feature, which makes sense given that teenagers make up a big chunk of the platform’s daily users. Teen Council member Sofia put it well: “Moving toward real-time rephrasing preserves the flow of conversation while still helping keep everyone safe. It gently guides the interaction back on track and helps people feel welcome.”

Now, is this AI fluff or genuinely useful? Honestly, this one feels like a real solution to a real problem. Hashtag walls in chat are genuinely annoying and make it harder to coordinate during gameplay. If the AI can rephrase a message in a way that preserves the meaning without the profanity, that’s actually better for everyone involved – the sender still gets their point across, and nobody has to squint at a row of pound signs trying to figure out what was said. That said, it’s worth watching how this plays out in practice. AI rephrasing is only as good as the model behind it, and there will inevitably be edge cases where the rephrased version changes the meaning in unexpected ways.

Alongside the new rephrasing tool, Roblox is also beefing up its core text filters. The company says it upgraded the underlying filter system to better catch variations of language designed to slip past Community Standards, including leet-speak, which is the trick where folks swap out letters for numbers or symbols to fool automated filters. Early results from testing show the improvements reduced false negatives for sharing or soliciting personal information, like social handles or phone numbers, by 20 times. That’s a notable jump in detection capability and a meaningful win for user safety on a platform with a large younger audience.

This update builds on work Roblox started in May 2025, when it introduced proactive, real-time warnings to prompt users to reconsider their language before hitting send. Experiments from last year showed that in-experience chat notifications and time-outs produced a 5 percent reduction in filtered chat messages and a 6 percent reduction in consequences from abuse reports. Small numbers individually, but meaningful at the scale Roblox operates at, where millions of people are chatting every single day.

Roblox is framing this as a first step on a longer path toward reducing hashtag censorship blocks for a more natural chat experience. The company said it will continue to iterate on these tools and plans to add additional controls for users down the road. For a platform trying to hold together a massive, age-diverse community, getting chat right matters more than it might seem – and this particular update looks like a genuine step in the right direction.

β˜•

Support independent tech journalism

NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.

Support NERDS.xyz
Avatar of Brian Fagioli
Written by

Brian Fagioli βœ”

Technology journalist and founder of NERDS.xyz

Brian Fagioli is a technology journalist and founder of NERDS.xyz. A former BetaNews writer, he has spent over a decade covering Linux, hardware, software, cybersecurity, and AI with a no nonsense approach for real nerds.