We develop efficient, differentially private algorithms for bilevel optimization with near-optimal excess risk bounds that nearly match the best rates for standard private optimization.