Moisture intrusions (MIs), episodes of extreme atmospheric moisture transport, significantly contribute to poleward moisture flux into the Arctic. At synoptic timescales, these events carry substantial moisture and heat, enhancing downward longwave radiation and triggering extreme warming events and sea ice loss. Over longer timescales, MIs potentially contribute to the projected amplified warming over the Arctic. Previous studies have identified large biases in the simulation of Arctic MIs in Coupled Model Intercomparison Project Phase 5 (CMIP5) models, which tend to underestimate MIs in the Atlantic sector and overestimate them in the Pacific sector. However, the causes of these biases remain unclear. In this study, we develop an automatic MI detection algorithm and apply it to CMIP6 models. Our findings reveal that the biases identified in CMIP5 persist into CMIP6. We further find that the underestimation of MIs over the Atlantic sector is attributed to an equatorward bias in the westerly jet over the North Atlantic. Conversely, overestimation in the Pacific sector results from a combination of a poleward bias in the upstream portion of the North Pacific jet and a weak bias in the downstream portion and at the jet core, both of which facilitate MIs propagating into the Arctic. Using additional model simulations, we demonstrate that correcting sea surface temperature biases and increasing atmospheric model resolution can partially alleviate these biases by improving the representation of the westerly jet. Our results suggest potential pathways to improve simulation of MIs, which, by improving simulation of the current Arctic climate, will lead to more credible projections of Arctic climate change.