How AI decision models could change content moderation

Decision models output a fixed set of choices instead of free text, which makes them faster and cheaper than large language models. Musubi's PolicyLM-1.7B applies a content policy written in plain English to messages in under 50 milliseconds and ships with open weights, so platforms can rewrite their rules without retraining it. That lets teams label content proactively at scale.

Summary written with AI by HowTechMoves, based on reporting by TechCrunch.