Artificial intelligence systems, particularly large language models (LLMs), have introduced a new paradigm of authority: one that operates without a subject, without reference to reality, and without epistemic validation. This paper introduces the Disconnected Syntactic Authority Theorem, which asserts that in generative language systems, authority can emerge solely from structural form. It is not derived from the truth of a proposition, the identity of a speaker, or the correspondence with the real, but rather from conformity to recognizable syntactic patterns. This structure-based authority performs commands, simulates expertise, and produces institutional tone—without intention, accountability, or consequence. The paper situates the theorem within critical discourse theory, functional linguistics, and algorithmic epistemology, showing how the autonomy of syntax becomes the foundation of a post-human legitimacy. The implications are profound: if authority no longer requires a body, then power becomes untouchable—functionally executed but politically invisible.