A Threshold on Responsibility in Human–AI Interaction

This document establishes a minimal threshold for responsibility in human–AI interaction,focusing on the risks introduced by persuasive and authoritative language generated byartificial systems. It does not propose a framework, methodology, or technical solution.It defines non-delegable conditions under which responsibility must remain humanwhen generative language systems are used in decision-bearing contexts. This publication serves as a boundary marker, not as an implementation guide.