Computing vibrational properties of crystals in the presence of complex defects often necessitates the use of (semi-)empirical potentials, which are typically not well characterized for perfect crystals. Here we explore the efficacy of a commonly used embedded-atomempirical interatomic potential for the U
x
Th1−x
O2 system, to compute phonon dispersion, lifetime, and branch specific thermal conductivity. Our approach for ThO2 involves using lattice dynamics and the linearized Boltzmann transport equation to calculate phonon transport properties based on second and third order force constants derived from the empirical potential and from first-principles calculations. For UO2, to circumvent the accuracy issues associated with first-principles treatments of strong electronic correlations, we compare results derived from the empirical interatomic potential to previous experimental results. It is found that the empirical potential can reasonably capture the dispersion of acoustic branches, but exhibits significant discrepancies for the optical branches, leading to overestimation of phonon lifetime and thermal conductivity. The branch specific conductivity also differs significantly with either first-principles based results (ThO2) or experimental measurements (UO2). These findings suggest that the empirical potential needs to be further optimized for robust prediction of thermal conductivity both in perfect crystals and in the presence of complex defects.