Credit scoring models play a pivotal role in financial institutions’ decision-making processes by assessing borrowers’ creditworthiness, and credit discipline, thereby mitigating credit risk. The credit scoring models used by financial institutions are based on transactional attributes like repayment history, sanctioned amounts, days past due and risk category status. This approach may not be sufficient to address the needs of all banking clients due to data collection inconsistencies, data absence and data governance and privacy issues. This paper proposes a novel approach to credit scoring models that leverages Generative Adversarial Networks (GANs), a module of Generative AI, for synthetic data augmentation. Traditional credit scoring models are often limited by data availability due to privacy concerns and imbalanced class distributions. GANs offer a solution by enabling the generation of realistic, synthetic customer data points that statistically resemble the true underlying data. By employing GANs, financial institutions can augment existing credit datasets, enhance model robustness, and improve predictive accuracy. The proposed approach involves training a GAN model on real credit data to capture the data distribution and subsequently generate artificial data samples. This synthetic data can address data scarcity issues, particularly for underserved populations, and mitigate the effects of imbalanced datasets commonly encountered in credit scoring. By generating synthetic minority class samples, the model’s ability to detect rare credit events is improved. This approach can vitalise financial institutions’ risk management practices and expand lending opportunities to previously unbanked segments of the population. Furthermore, GAN-generated synthetic data helps protect sensitive customer data, promoting responsible lending practices. Overall, this paper highlights the potential of GANs to create more accurate, dependable, and inclusive credit risk models.