An organization is collecting data to train a generative AI model for customer service. They want to ensure security throughout the ML lifecycle. What is a critical consideration at this stage?