AI Limits: How Nations Could Cooperate & Exit Agreements

Researchers propose a framework for international agreements designed to limit the development of frontier AI and outlined conditions for relaxing those limits. They examined existing international agreements to identify properties appropriate conditions should satisfy. The study suggests a fixed time period for limitations, after which a new organization would define conditions for safe AI development.

The authors acknowledge determining when to lift restrictions presents a challenge. They propose allowing withdrawal from the agreement under extraordinary circumstances, recognizing the need for flexibility.

This research aims to illustrate the complex considerations involved in creating an effective international AI governance system. The framework could inform future negotiations and help nations collaborate to manage the risks associated with increasingly powerful artificial intelligence.