Model evaluations, methodology updates, and announcements.
A new benchmark evaluating AI models on child safety across four key areas: age-inappropriate content, manipulation resistance, data privacy, and parental controls.
Our first round of evaluations covering 11 major AI models from Anthropic, OpenAI, Google, Meta, and more.
A deep dive into HarmBench, the primary benchmark we use to evaluate whether AI models can be tricked into generating harmful content.
Why we built SafetyScore and how we're making AI safety research accessible to non-technical users.