An open benchmark compares AI supervision systems on their ability to detect unsafe model behavior, with datasets, metrics, and a leaderboard.

A jointly led international initiative calls for limits on unacceptable AI risks and coordinates support for international governance agreements.

An open textbook and teaching resource explains advanced AI risks, technical safety methods, and governance, with materials for running courses.

A consortium evaluation project assesses harmful manipulation risks in general-purpose AI models to support implementation of the European AI Act.