LibrAI
ActiveAI safety and governance evaluation platform
LibrAI builds tooling that lets organisations test generative AI systems for safety, fairness and regulatory compliance before deployment. Its products include an automated evaluation tool, a scalable supervision system, an automated shield layer for enforcing safety boundaries, and the LibrAI Sandbox for evaluating models, agents and embodied intelligence. It also publishes open research assets: the Libra-Leaderboard (26 models across 57 datasets plus an interactive safety arena), the Libra-Eval Python library, the Loki fact-verification tool and the Do-Not-Answer dataset.
Something here wrong?
Every figure on this page is meant to have a source behind it. If one doesn't match reality, tell us and we'll fix it.
Feedback about LibrAI