Additive manufacturing (AM) offers unprecedented design freedom for jet engine components, yet its qualification remains challenging due to inherent process-induced defects. Among these, lack-of-fusion (LoF) flaws are particularly critical: although rare, they pose a risk to structural integrity due to potential grow under cyclic loading and high-temperature conditions. Non-destructive inspection (NDI) systems (e.g., X-ray) censor the defect population below a size threshold with a given probability of detection (PoD). Additionally, this can vary across the part geometry. Moreover fracture mechanics analyses reveal that critical defect sizes may fall below the detection threshold, creating a safety gap that deterministic design cannot fully address. This work presents a probabilistic framework for robust design under such uncertainty that explicitly models this gap. Building on extreme value statistics of defect populations, as pioneered by Beretta and others [1], the distribution of defect sizes and frequencies is modeled to quantify the residual risk associated with undetectable flaws. The methodology integrates defect statistics, POD curves, and fracture mechanics to estimate failure probabilities and provide design margins. Furthermore, several approaches to decrease the amount of necessary computations by reducing uncertainties with conservative assumptions are discussed. A representative case study demonstrates how the proposed framework supports robust design decisions and highlights the importance of combining statistical defect characterization with reliability analysis and NDI capabilities to ensure structural integrity beyond deterministic limits. REFERENCES [1] S. Beretta, More than 25 years of extreme value statistics for defects: Fundamentals, historical developments, recent applications, International Journal of Fatigue, Volume 151, 2021, 106407, ISSN 0142-1123, https://doi.org/10.1016/j.ijfatigue.2021.106407.