The manufacture of defect-free parts has been a key discussion topic with the widespread adoption of additive manufacturing by industry. While significant research has been performed on the detection of powder bed defects, the focus has been on the classification of the defects according to defect type. However, when looking at creating a closed loop feedback system, it is important for the machine to make autonomous decisions regarding defects. The focus of this paper will be to create a defect severity classification matrix based on industry partner experience as well as published literature that can be used to autonomously classify defects