Overview
- Roblox introduced real-time rephrasing that converts flagged words into more respectful wording to keep chats readable, replacing the old “####” output.
- Participants see a notice when a message is rephrased, and the wording stays close to the sender’s original intent.
- Rephrased messages still count as policy violations, and users who continue to curse face the same enforcement actions.
- The feature is limited to in-experience chats between age-verified users in similar age groups, supports all languages available through automatic translation, and was built with input from the Teen Council.
- Roblox upgraded its filtering stack with an additional AI layer and larger reasoning models, citing better detection of leetspeak and a 20x reduction in false negatives for personal-information sharing.