This study aims to investigate the impact of using Artificial Intelligence and Machine Learning on gender bias in credit scoring models by comparing advanced estimation techniques (Random Forest, Support Vector Machine, Artificial Neural Networks) with traditional methods (logistic regression). As AI-based credit scoring systems become widespread, concerns about transparency, fairness, and potential discrimination arise, especially regarding sensitive attributes like gender. Using data from the National Bank of Romania's Credit Risk Register, this study spans a seven-year period, offering an empirical analysis of potential biases in mortgage lending. Findings indicate that, while ML models provide enhanced predictive power, they vary in fairness. Random Forest emerges as the most accurate and least discriminatory model, underscoring the need for careful model selection to ensure equitable credit decisions.