Do AI models look before they act?
A new benchmark finds frontier AI models often skip checking for safety information before acting, with sharp differences across models.
A new benchmark finds frontier AI models often skip checking for safety information before acting, with sharp differences across models.