We are now the AI Security Institute
Please enable javascript for this website.

Safety case template for frontier AI: A cyber inability argument

Read the Full Paper

Authors

Marie Buhl
Tomek Korbak
Jessica Wang
Benjamin Hilton
Geoffrey Irving

Marie Buhl, Tomek Korbak, Jessica Wang, Benjamin Hilton, Geoffrey Irving

Abstract

Introduces a safety case template for AI with offensive cyber capabilities, providing a structured way to assess and argue acceptable risk through evidence and clear claims. It aims to advance AI safety assurance by integrating key safety techniques into a coherent framework.

Notes