AI models revolutionize content moderation with new open-source tool

Musubi, a company specializing in AI decision models, has introduced PolicyLM-1.7B, an open-source model designed for real-time content moderation. The model, released on October 6, 2026, can apply content policies written in plain English to messages in under 50 milliseconds. It is intended to be as efficient as existing AI classifiers used by social platforms but with the flexibility of a modern large language model. PolicyLM-1.7B does not require retraining when policies change, allowing for rapid policy updates. Filip Jankovic, Musubi’s co-founder and chief AI officer, highlighted the model’s ability to help platform managers proactively label content, especially as content volume grows. He noted that product teams seek better insights into their platforms. Jankovic’s interest in decision models dates back to a 2024 project called GLiNER. The company is leveraging the growing interest in decision models to promote transparency in content moderation, positioning PolicyLM-1.7B as a tool that users can run themselves. Decision models, which output outcome probabilities rather than text, have gained traction since the release of Typesafe AI’s Jev in September 2025. Competitors like OpenAI and Amazon have also developed similar models. These models are faster and cheaper than large language models while maintaining transformer architecture flexibility. One use case is monitoring AI agent behavior, with human misbehavior being a natural extension.