Posterior contours of the seven cosmological and astrophysical parameters which are varied given the anisotropic systematics model within KiDS-SBI over the prior space shown in Table 2 when analysing a mock data vector from the same model. The black solid lines indicate the true cosmology of the input mock data vector generated from the anisotropic systematics model while adding noise. All the aforementioned values are shown in Table E.1. These posteriors are obtained from training neural density estimators in DELFI (Alsing et al. 2019) on 18,000 realisations of the forward simulations assuming the anisotropic systematics model, in line with the choices made in L23. The posterior is obtained from the combined posteriors of six independent conditional Masked Autoregressive Flows (MAFs) each is made up of three to eight Masked Autoencoder for Density Estimations (MADEs) each with two hidden layers of 50 neurons.