You have implemented a custom computer vision model using Azure AI Vision and deployed it to production. Over time, you notice that the model's performance degrades, and it starts to produce more incorrect predictions. What could be the possible reasons for this?