Artificial intelligence is increasingly determining loan approvals and credit assessments in the personal banking sector, often without human oversight. While these algorithms are marketed as neutral and efficient, they frequently perpetuate gender bias because they rely on historical data that fails to accurately reflect women's financial realities. A recent report by the EU Agency for Fundamental Rights highlights a significant gap between legal ambitions and practical implementation regarding high-risk AI systems. Providers often lack the tools or expertise to systematically assess discrimination risks, leading to inconsistent self-assessments and weak oversight. When data inputs do not capture the depth of women's lives, the resulting algorithms reproduce existing social and economic disparities. Systems operate by identifying patterns in historical data and projecting them forward. If women are underrepresented or their outcomes are not analyzed separately from men, the system cannot detect unequal treatment. This creates a state of algorithmic blindness where inequality is embedded silently. A notable example from Kenya illustrates this issue: a digital lending algorithm consistently offered women smaller loans than men, sometimes by over a third, despite women demonstrating stronger repayment performance. The system did not intentionally discriminate; it simply mirrored the historical biases present in the data it was trained on. To overcome these biases, the implementation of sex-disaggregated data is essential. While most financial institutions record customer gender for identification purposes, this information is often collected but never analyzed, reported, or monitored. In many jurisdictions, this variable is filed away and ignored by supervisors, leaving gender gaps hidden and becoming permanent features of the financial landscape. In contrast, some developing nations have successfully integrated this data into their regulatory frameworks. In Latin America, regulators have long required and published sex-disaggregated financial statistics. In Mexico, authorities combine bank data with household surveys to analyze how different genders use financial services. This visibility led to concrete policy changes; for instance, supervisors discovered that women's loans were smaller but less risky, prompting adjustments to loan loss provisioning rules. Similarly, in Chile, tracking gender data revealed that equal access to accounts did not result in equal outcomes for savings or insurance, leading to more targeted policy responses. These examples suggest that the lack of gender data in high-income economies, particularly in Europe, stems from institutional hesitation rather than technical limitations. As the EU implements the AI Act, the absence of systematic gender data poses a fundamental challenge: how can fairness be monitored if the metrics needed to detect inequality are never analyzed? Making women visible in data is not merely symbolic but a practical prerequisite for fair finance. Without sex-disaggregated data, claims of algorithmic fairness remain unsubstantiated, and automated systems continue to replicate historical injustices at scale.