The development of reliable additive manufacturing (AM) technologies to process metallic materials, e.g. selective laser melting (SLM), has allowed their adoption for manufacturing final components. To date, ensuring part quality and process control for low-volume AM productions is still critical because traditional statistical techniques are often not suitable. To this aim, extensive research has been carried out on the optimisation of material properties of SLM parts to prevent defects and guarantee part quality. Amongst all material properties, defects in surface hardness are of particular concern as they may result in an inadequate tribological and wear resistance behaviour. Despite this general interest, a major void still concerns the quantification of their extent in terms of probability of defects occurring during the process, although it is optimised. Considering these issues, this paper proposes a novel approach to quantify the probability of occurrence of defects in hardness-optimised parts by SLM. First, three process variables, i.e. laser power, scan speed and hatching distance, are studied considering their effect on hardness. Design of Experiments and Response Surface Methodology are exploited to achieve hardness optimisation by controlling process variables. Then, hardness defect probability is estimated by composing the uncertainty affecting both process variables and their relationship with the hardness. The overall procedure is applied to AlSi10Mg alloy, which is relevant for both aerospace and automotive applications. The approach this study proposes may be of assistance to inspection designers to effectively and efficiently set up quality inspections in early design phases of inspection planning.