This paper reduces the gap between lower and upper bounds for fixed-confidence Best Arm Identification under global Differential Privacy to a small constant multiplicative factor by introducing a new lower bound and a near-optimal Top Two algorithm.