In recent years there has been a lot of interest in F-Theory GUTs, mostly considering local models. In this talk I first consider an SU(5) GUT locally at the "point of E_8". The requirements of proton stability and reasonable flavour structure single out exactly two models. However, both models cannot be embedded in a semilocal construction (via the spectral cover approach). This casts doubts on the predictivity of local models.