Multi‑agent language‑model systems often add a reviewer to improve an upstream agent's decision. A reviewer can both fix errors and replace decisions that were already correct. We formulate this as an institutional jurisdiction problem: given the same proposal, task information, model, and review output, when is the reviewer’s recommendation binding? We compare two authorization schemes: a broad override authority and an evidence gate that permits replacement only after the reviewer produces a mechanically valid record of a binding‑constraint violation and a compliant repair. Experiments run on two locally hosted open‑weight models (Qwen and Mistral), 96 spatial‑policy tasks, and a 96‑task portfolio‑transfer benchmark. The gate sharply reduces corruption of correct proposals while retaining most violation repairs. The direction of this trade‑off is partly implied by the rule; its magnitude depends on reviewer transitions and upstream proposal states. A prospective validation lets the reviewer explicitly choose to retain, replace, or escalate. This eliminates correct‑proposal corruption for Qwen, but Mistral still replaces all 48 correct proposals; Qwen also declines every welfare‑only repair that Mistral frequently attempts. Thus interface design changes reviewer behavior but is not a model‑independent safeguard. The result is not that restricted review is universally better, but that reviewer competence, action interface, and binding jurisdiction jointly determine both the correction domain and the corruption surface.
Review