Material strengthening and embrittlement are controlled by intrinsic interactions between defects, such as grain boundaries (GBs), and impurity atoms that alter the observed deformation and failure mechanisms in metals. In this work, we explore the role of atomistic-scale energetics on liquid-metal embrittlement of aluminum (Al) due to gallium (Ga). Ab initio and molecular mechanics were employed to probe the formation/binding energies of vacancies and segregation energies of Ga for 〈100〉, 〈110〉 and 〈111〉 symmetric tilt grain boundaries (STGBs) in Al. We found that the GB local arrangements and resulting structural units have a significant influence on the magnitude of the vacancy binding energies. For example, the mean vacancy binding energies for 〈100〉, 〈110〉 and 〈111〉 STGBs in the 1st layer was found to be −0.63, −0.26 and −0.60eV, respectively. However, some GBs exhibited vacancy binding energies closer to bulk values, indicating interfaces with zero sink strength, i.e. these GBs may not provide effective pathways for vacancy diffusion. The results from the present work showed that the GB structure and the associated free volume also play significant roles in Ga segregation and the subsequent embrittlement of Al. The Ga mean segregation energies for 〈100〉, 〈110〉 and 〈111〉 STGBs in the 1st layer were found to be −0.21, −0.09 and −0.21 eV, respectively, suggesting a stronger correlation between the GB structural unit, its free volume and the segregation behavior. Furthermore, as the GB free volume increased, the difference in segregation energies between the 1st layer and the 0th layer increased. Thus, the GB character and free volume provide an important key to understanding the degree of anisotropy in various systems. The overall characteristic Ga absorption length scale was found to be about ∼10, 8 and 12 layers for 〈100〉, 〈110〉 and 〈111〉 STGBs, respectively. In addition, a few GBs of different tilt axes with relatively high segregation energies (between 0 and −0.1eV) at the boundary were also found. This finding provides a new atomistic perspective for the GB engineering of materials with smart GB networks to mitigate or control liquid-metal embrittlement and more general embrittlement phenomena in alloys.