We consider the problem of learning linear dynamical systems from input-output data, or system identification, given limited output samples. Learning the dynamics is often the basis of control or policy decision problems in tasks varying from linear-quadratic control to deep reinforcement learning. Recent literature provides finite-sample statistical analysis using least-squares regression. When a low-order system (corresponding to a low-rank Hankel matrix) is desired, adding a Hankel nuclear norm regularizer is common in engineering practice, but has had unknown sample complexity. In this talk, we present a finite-sample analysis and new insights for system identification with this regularized scheme.