Enforcement, Not Attribution: An Empirical Feasibility Envelope for AI-Assisted Cyber-Weaponization Governance

AI-Assisted Cyber Weaponization (AIACW) is emerging as a society-scale risk. The Pharos Lighthouse proposal (UN CTED submission, April 2026) argues for a coordinated global-defense architecture as response. The proposal has been critiqued on empirical grounds: is the proposed architecture feasible under endogenous adoption, and is it effective against the status-quo no-coordination baseline We address both questions with a 5000-episode Monte Carlo simulation of coalition dynamics across 30 heterogeneous jurisdictions, comparing three regimes (no coordination / partial coalition / full Pharos) under bounded-rational adoption on a Watts-Strogatz peer network. Under the pre-registered parameter setting, the full-Pharos regime reduces aggregate harm against the no-coordination baseline by −1.13 95% bootstrap CI: −1.18, −1.09, with the coalition stabilizing at approximately 17 of 30 jurisdictions; in 0 of 5000 episodes does the full-Pharos regime produce more harm than no-coordination. A parameter sweep across attribution success, free-rider temptation, peer influence, and enforcement strength confirms that S2 dominates S0 on aggregate harm across the full envelope tested; enforcement strength is the dominant lever, attribution success and peer influence are sub-dominant by more than an order of magnitude. We pair this simulation result with a review of single-jurisdiction cyber-enforcement outcomes 2010–2024, which display the same ordering: attribution has advanced substantially while aggregate harm has grown, a trajectory consistent with the structural prediction of the model. The contribution is a reproducible feasibility envelope, not a geopolitical prediction. This paper is one of two empirical anchors for the parent record Pharos Lighthouse (Zenodo DOI 10.5281/zenodo.19645912). The companion paper P7 (Residual Manipulability of Frontier Aligned Language Models) addresses the threat-side empirical question. All code, parameters, and random seeds are deposited under the AIACW Empirical Teasers concept DOI 10.5281/zenodo.19687373 (current v1.0: 10.5281/zenodo.19771546); CC-BY-4.0 (paper) / MIT (code).