BKC Faculty Associate Arvind Narayanan and Sayash Kapoor argue that current safety efforts are limited because AI safety is not a model property. "Trying to make an AI model that can’t be misused is ...