According to DeepLearning.AI, a recent study found that OpenAI’s GPT-4o was able to accurately identify verbatim text from both public and paywalled O’Reilly Media books, indicating the model was likely trained on copyrighted content (source: DeepLearning.AI, June 7, 2025). This finding raises potential legal and ethical concerns for AI model training practices, which could impact AI-related tokens and projects in the cryptocurrency market due to increased regulatory scrutiny and potential lawsuits.