Lingua Pactum Protocol: Constitutional Admissibility for AI-Triggered Language and Action

This preprint introduces the Lingua Pactum Protocol (LPP), a constitutional admissibility framework for AI-triggered language and action. LPP distinguishes runtime governance from upstream constitutional admissibility. Runtime governance determines whether an already-formed request should be allowed under policy. Constitutional admissibility asks whether an AI-triggered statement or action is qualified to become an execution candidate in the first place. The paper argues that policy compliance is insufficient for constitutional admissibility, and that runtime governance cannot legitimize itself without an upstream admissibility layer. It introduces a minimal admissibility model, defines admissibility collapse, and positions LPP relative to recent work on pre-action legitimacy, judgment-gated automation, tool-call authorization, runtime governance, policy engines, workload identity, and authorization token systems.