As decision models spread across the industry, a company called Musubi has a brand new idea for how put them to work: moderating content. On Tuesday, Musubi introduced a lightweight decision model made for real-time moderation called PolicyLM-1.7B, launched with open weights.
The idea is to take a content policy written in plain English and apply it to messages in under 50 milliseconds. Musubi’s model is designed to be similar in cost and speed to the AI classifier systems that power moderation on most social platforms — but as it has the ability of a modern LLM, it can apply complex policies without special training. Even more vital, the model won’t require new training whilst the policy changes, permitting for human policy-setters to iterate as much as they want.
As Musubi co-founder and leader AI officer Filip Jankovic sees it, it gives platform managers a way to label content proactively.
“Product teams just want a better understanding of what’s happening on their platform, particularly as the amount of content is exponentially growing,” Jankovic says. “Being able to label all of that in a very scalable, customizable way is extremely beneficial.”
Decision models have become a hot topic in the AI world since the launch of TypeSafe AI’s Jev in September, which was rapidly followed by competing decision models from OpenAI and Amazon. Rather than of outputting text content, a decision model outputs final results possibilities, though in this situation the model outputs a binary judgement: Either the content is in the category or it isn’t. By restricting the model’s output to a set of predetermined choices, decision models are capable of run faster and less expensive than large language models, even as maintaining the flexibility of the transformer architecture.
One early use case is reining in misbehavior by AI agents — so it’s only natural to apply the same technology to human misbehavior.
Notably, Jankovic says his interest in interest models predates Jev, tracing it back to a 2024 project called GLiNER (Generalist Model for Named Entity Recognition) that deployed many of the same strategies.
Still, Musubi isn’t wary of the comparison. If anything, the company is eager to use the new interest in decision models to shine a light on content moderation. “If Jev caught your eye, PolicyLM-1.7B is the same form of model, trained mainly for content moderation, that you can run yourself,” the product declaration reads.
