An AI risk assessment reveals that a fraud detection model's performance varies significantly depending on which data engineer prepared the training dataset. Upon investigation, the team finds no standardized data preprocessing pipeline. What category of AI risk does this MOST represent?

B is correct because when model outcomes depend on ad hoc, undocumented data preparation decisions, the model cannot be reliably reproduced, validated, or audited. This is model reproducibility risk: if the same model is retrained by a different engineer, it may produce materially different results because the preprocessing is not standardized. This undermines model governance and validation integrity. Algorithmic bias (A) is a potential downstream effect but not the root risk identified. Data privacy (C) is a separate concern. Operational risk (D) is a related concern but secondary to the model risk.