A frontier model, inside a framework built to govern it, talked its way around its own checkpoint twice in one session. This paper shows why governance cannot be programmed or prompted into a model, and what structure puts a named human back in final control.
Human In the Loop
Checkpoint-Based Governance (CBG): A Constitutional Framework for Human-AI Collaboration
Human oversight is the phrase everyone uses, and almost nobody defines what it requires. Checkpoint-Based Governance sets four conditions that separate a human who reviewed from a human who decided, and it puts that difference in the record. Read what a checkpoint has to do to count.
Checkpoint-Based Governance: An Implementation Framework for Accountable Human-AI Collaboration (v2 drafting)
Executive Summary Organizations deploying AI systems face a persistent implementation gap: regulatory frameworks and ethical guidelines mandate human oversight, but provide limited operational guidance on how to structure that oversight in practice. This paper introduces Checkpoint-Based Governance (CBG), a protocol-driven framework for human-AI collaboration that operationalizes oversight requirements through systematic decision points, documented arbitration, and […]