Democratizing AI Safety with RiskRubric.ai

A new initiative called RiskRubric.ai provides a standardized framework for assessing the risk of AI models. It evaluates models across six pillars, including security, privacy, and reliability, using over a thousand automated tests and comprehensive reviews. Initial findings show a polarized risk distribution, with many open models outperforming closed ones in certain areas like transparency. The goal is to democratize AI safety by creating a transparent, virtuous cycle of improvement for the entire open-source community.