Contract monitoring: governing AI via separation of powers.

arXiv:2609.32061v1 Announce Type: new Abstract: We propose an AI safety framework that binds worker agents to contracts specifying their permitted actions. We show how these contracts can be enforced and specified by assigning distinct responsibilities to moni