According to DeepLearning.AI, a recent study revealed that OpenAI’s GPT-4o has likely been trained on copyrighted, paywalled content from O’Reilly Media books. Researchers evaluated GPT-4o and other leading AI models by testing their ability to identify verbatim text from both public and private book excerpts. The findings indicate that GPT-4o was able to accurately reproduce content from paywalled O’Reilly books, suggesting potential copyright and licensing issues for AI training datasets. This has significant implications for AI industry practices, particularly in compliance, data sourcing, and the development of future large language models. Businesses relying on AI-generated content may need to reassess their risk management strategies and ensure proper licensing, while AI developers face increasing pressure to adopt transparent data curation methods (Source: DeepLearning.AI, June 7, 2025).