Can Humans Devise Practical Safeguards That Are Reliable Against an Artificial Superintelligent Agent?

Dramatic advances in frontier artificial intelligence (AI) raise the question of whether humans can design practical safeguards to protect our critical and digital infrastructure against attacks from a future artificial superintelligence. Some argue that a superintelligent agent would have such mastery and power over physical and cyber space that it could thwart any effort by humans to prevent it from undermining the confidentiality, integrity and availability of our information. Others recognize the possibility that we might design security that is rooted in fundamental limits that even a superintelligence could not overcome. 

In this paper, the authors present a hypothesis that such safeguards are both feasible and sensible, outline an approach rooted in existing security practice, and call for future work that enlists security engineers and AI developers in common pursuit of these safeguards. Through an example protocol and threat model, the authors explore how existing and adapted security tools can meaningfully restrict the actions of advanced AI, while acknowledging the limitations and assumptions inherent in any security protocol. They conclude by outlining directions for future research, emphasizing the need for rigorous formalization and ongoing evaluation as both AI capabilities and the technical environment evolve.