Algorithmic bias is governed by Article 10(2)(f) of the AI Act which requires providers of high-risk systems to examine training, validation and testing data for biases likely to affect health, safety or fundamental rights. The obligation is about data governance, not a general fairness principle.