Large language models have emerged as the new lingua franca of geopolitical power — mediating communication across languages and cultures while embedding, within that very mediation, the governance preferences, content norms, and political values of whoever built, trained, and regulated them. This dual character — simultaneously a communication infrastructure and a normative architecture — makes LLM governance the most consequential frontier of the current AI regulatory contest. That contest is most commonly framed as a binary choice between China’s state-led, security-first model and the European Union’s rights-based, risk-categorised framework. Yet the most revealing laboratories of this rivalry are neither Washington nor Brussels nor Beijing, but Riyadh, Abu Dhabi, and Cairo. Gulf Cooperation Council (GCC) states, Egypt, and Iran are simultaneously receiving Chinese-built Arabic LLM ecosystems and Western foundational model platforms under governance frameworks that belong fully to neither camp. This paper examines how Middle Eastern states exercise agency in adopting, filtering, or independently constructing LLM governance norms — asking whether they are passive recipients of a China-West regulatory binary, active norm-brokers, or emergent norm-makers in their own right.