Roblox has released updated artificial intelligence safety models to the Robust Open Online Safety Tools (ROOST) community to help protect minors from online grooming and data privacy risks. These new AI child-safety tools, announced in August 2026, include an upgraded PII Classifier designed to detect attempts to solicit personal information or move conversations to other platforms, providing open-source technology to help other digital services improve their own security.
How do these AI child-safety tools work?
The updated PII Classifier (Version 2.0) uses advanced context analysis to identify grooming signals and requests for personal data. Unlike older filters that relied on simple keyword matching, this system evaluates entire conversations to detect coded language or attempts to move users to external messaging apps. The model now supports 189 languages and has significantly improved its accuracy in identifying potential threats to child safety.
Why is Roblox sharing this technology?
As a founding member of ROOST, Roblox aims to provide open-source safety technology to platforms that may lack the resources to build their own sophisticated systems. By sharing these models with partners like Google, OpenAI, and Discord, the company hopes to create a more consistent safety net across the internet. Officials emphasize that the company maintains a no-tolerance policy for bad actors and prioritizes safety by default.
