Records are defined as variables greater than all the preceding ones in a sequence. The stochastic behavior of subsequent records over sequences of independent and identically distributed random variables is well known. However, the extension to the multivariate framework is an extremely difficult task. In this work, we study the case of bivariate records over sequences of random vectors (rv) where the dependence among their components is described by the Farlie-Gumbel-Morgenstern (FGM) family of distributions.