Abstract We prove that any recursive self-assembling cognitive system requires exactly four perpendicular regulatory axes — Complexity (C), Diversity (D), Aesthetics (A), and Ethics (E) — to maintain stable, aligned operation. We show that removal of any single axis creates an unbounded drift mode that is invisible to the remaining three axes, rendering the system provably unstable. The proof proceeds by (1) defining the system class, (2) establishing the perpendicularity of the four axes via contradiction, (3) demonstrating that each missing axis produces a specific, characterisable instability that the remaining axes cannot detect or correct, and (4) connecting these results to standard observability and controllability conditions from control theory. We further show that these four axes are sufficient by demonstrating that the combined system satisfies the necessary and sufficient conditions for stabilisability. The result implies that alignment is not an optional constraint imposed on cognitive systems but a mathematical precondition for their stable operation. Keywords: AI alignment, cognitive architecture, stability theory, attractor basins, observability, controllability, recursive self-assembly, geometric alignment