In recent years, the integration of new Artificial Intelligence (AI) techniques and capabilities has emerged as one of most promising research fields to aid the industrial development of smart and zero-defect manufacturing solutions. This study explores the potential of generative AI in this field and reviews novel opportunities enabled by generative AI methods, and Generative Adversarial Networks (GANs) in particular, to aid the generation of augmented datasets including realistic representations of anomalous process patterns. The result is an effective AI framework to learn specific defect features from real data, and reproduce them in an extended way, leading to synthetic but realistic image data that could be used to enhance defect detection and classification performances. The paper reviews the benefits and open challenges associated with the implementation of these techniques, including state-of-the-art examples and real case studies in Additive Manufacturing.