A retail company is evaluating a third-party AI recruiting tool that its AI council must approve before deployment. During review, the council discovers the vendor cannot explain how the model produces candidate rankings and provides no documentation on the training data or decision logic. The council wants to ensure the deployed solution meets Microsoft's responsible AI standards. Which responsible AI principle is most directly at risk given the vendor's inability to explain the model's outputs?