In a recent paper, our team demonstrated that large language models (like the ones behind ChatGPT) can be used to train algorithms that predict fraud-related securities class action lawsuits.