The number of HIV and AIDS cases is a count data correlated with each other. The common model to regress two response variables in the form of count data is bi-response Poisson regression. One infraction of assumptions in the Poisson regression is over dispersion. Negative binomial regression is to be solution to this over dispersion case. This study aims to estimate bi-response negative binomial nonparametric regression model based on local linear estimators for modeling the number of HIV and AIDS cases in East Java. The results give the deviance value as the goodness of fit criterion of bi-response negative binomial nonparametric regression model by using local linear estimator of 2.263 which is smaller than the deviance value of the parametric regression model of 3.926. It means that the bi-response negative binomial nonparametric regression model by using local linear estimator is better than that by using parametric regression model approach to explain the relationship between HIV and AIDS cases affected by drug users in East Java.