Treaty-Following AI

Maas and Olasunkanmi propose agents that follow their principals’ instructions except where the goals or actions would breach a designated AI-guiding treaty. Their framework would make treaty compliance part of the agent’s operating constraints.^[1]

The proposal is conditional on unresolved technical, legal, and political problems. It does not establish that current agents can reliably interpret or enforce treaties. This places AI Alignment alongside international institutions and Social Alignment. See Treaty-Following AI and Treaty-Following AI.