Closing the Gap Between Policy and Practice

Runway has updated its content moderation system to better distinguish between creative expression and real harm. By leveraging advances in large language models (LLMs), the platform can now apply more nuanced, custom policies instead of relying on overly broad third-party classifiers. This change allows for the creation of content involving conflict and horror, such as sword fights or shootouts, which were previously often blocked. While the new system is more accurate, limitations remain, including mistakes by automated systems and moderation applied by integrated third-party models. Runway encourages users to report incorrect moderation decisions to help further refine its policies and tools.