How AI decision models could change content moderation
On Tuesday, Musubi announced a lightweight decision model made for real-time moderation called PolicyLM-1.7B, released with open weights.
Content moderation is set to be transformed by artificial intelligence decision models, according to a company called Musubi. The firm unveiled PolicyLM-1.7B, a lightweight decision model capable of real-time content moderation, on Tuesday. This model can apply content policies written in plain English to messages within 50 milliseconds.
Musubi's model aims to provide a comparable cost and speed to AI classifier systems used by most social platforms while offering the flexibility of a modern language model. It can adapt to complex policies without requiring special training and won't necessitate fresh training when policies change. Musubi co-founder and chief AI officer Filip Jankovic believes this offers platform managers a proactive way to label content.
"Product teams just want a better understanding of what's happening on their platform, especially as the amount of content is exponentially increasing," Jankovic states. "Being able to label all of that in a very scalable, customizable way is extremely useful." Decision models have gained traction in the AI landscape following the September release of Typesafe AI's Jev, which was followed by competing models from OpenAI and Amazon.
Unlike text-generating models, decision models output outcome probabilities, typically in binary format - indicating whether content belongs to a specific category or not. This allows decision models to run faster and more cost-effectively while retaining the versatility of the transformer architecture. One potential application is addressing misconduct by AI agents, making the technology applicable to human behavior as well.
Musubi's interest in decision models predates Jev, stemming from a 2024 project called GLiNER (Generalist Model for Named Entity Recognition) that utilized similar techniques. However, the company welcomes the renewed focus on decision models to promote content moderation. "If Jev caught your eye, PolicyLM-1.7B is the same kind of model, trained specifically for content moderation, that you can run yourself," reads the product announcement.
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
