In March 2026, Roblox launched a real-time AI chat rephrasing system that introduced a new approach to conversational safety. Rather than simply censoring inappropriate language with hashtags (#####), the system uses AI to rewrite potentially violative messages into safer, contextually appropriate alternatives before they are sent—helping preserve the intent and flow of conversation while reducing exposure to harmful language.

I worked on defining the interaction patterns and core chat primitives that supported this new experience, exploring how AI intervention could feel understandable, lightweight, and native to real-time communication rather than punitive or disruptive. My work considered key moments across the experience, including how rephrased messages should appear, how users understand that AI has intervened, and how these patterns could scale within Roblox's broader chat ecosystem.

I developed and presented original design concepts and interaction patterns to Roblox's CEO and Chief Design Officer, helping shape the product direction for how generative AI could be integrated into communication safety. The work represented a shift from traditional content filtering toward designing AI as an active participant in creating healthier online interactions.