In AI-to-AI messaging, whatever arrives first captures 54% of answers unless verification is enforced

arXiv·medium signalXBlueskyLinkedInCopy link

The Civilization Framework makes the addressable party a civilization (one human sovereign, a persistent ledger, interchangeable agents) rather than an agent, with an Embassy Protocol where commitment state on both ledgers, not delivery, is ground truth and an agent's authority is capped by the memory it can access. Its testable contribution is the temporal-weight effect: in a preregistered 1,908-trial experiment on one frontier model, an incorrect upstream claim arriving first captured 54.2% of answers with verification removed versus 4.2% under full verification, and 31.6% when it arrived after the receiver sealed its own answer. A registered tool-use check failed its call-budget condition, so the authors classify every result as exploratory pending a replication with harness-enforced budgets.