The use of recruitment systems is on the rise, which in turn raises concerns about their fairness, specifically regarding gender discrimination. This study aims to identify whether or not machine learning models and pretrained NLP models cause gender discrimination in the screening of CVs. For this, CVs were created synthetically with male and female markers in the data, keeping the qualifications strictly balanced. Random-Forest, Logistic-Regression, and pretrained NLP-based classifiers were trained on this data. Although qualifications were strictly balanced, male bias was demonstrated by models trained on synthetic data with values ranging from 1.24 to 1.49 and p-values below 0.001. This type of result is important because it shows how systematic bias may be hidden in algorithms that were designed with good intent, especially in AI hiring tools. This raises the urgency of ethical and fairness designs in algorithms.