AI Training Data Quality Risk arises when datasets used to develop machine learning models contain errors, biases, gaps, or fail to represent the population where the model will be deployed. Low-quality training data leads to models that perform poorly in production, make biased decisions, or fail unexpectedly in edge cases. For example, facial recognition systems trained predominantly on lighter-skinned individuals demonstrate higher error rates for people of color, creating discrimination risks. Organizations must implement data quality controls, bias testing, diversity assessment, and ongoing monitoring of model performance across demographic segments. The EU AI Act explicitly addresses data governance requirements for high-risk AI systems.