Reddit rolling out AI bouncer to halt harassment
Photo by Brett Jordan
Brandon Vigliarolo | The Register
Reddit is implementing a new Large Language Model (LLM)-powered harassment filter to aid its volunteer moderators in combating online harassment, discovered in a recent APK teardown of the Reddit app for Android. This optional community safety tool allows moderators to automatically filter posts and comments deemed harassing, with the LLM trained on previous moderator actions and content removed by Reddit's internal enforcement teams. The harassment filter offers settings for varying degrees of content filtering, from low to high, and includes an explicit allow list for bypassing the filter for certain keywords.
As Reddit prepares for its Initial Public Offering (IPO), it has been actively incorporating AI into its platform, including a deal allowing Google to access its Data API for training models and enhancing Reddit's search capabilities. The introduction of AI moderation tools and other AI integrations come amidst concerns from Reddit's community about the platform's direction and potential overreliance on technology, reflecting broader challenges as the company moves towards going public.