OpenAI has released an update to its GPT-5 system card, detailing the latest model family in the series. The safety mitigation approach for these models remains largely unchanged from previous versions. GPT-5.2 models were trained on a diverse range of datasets, including publicly available internet information and data provided by users, third-party partners, or human trainers. --- Why it matters: This update is relevant to researchers working with large language models, as it provides insight into the training data and safety measures used in the GPT-5.2 series. Understanding these factors can inform the development of more robust and responsible AI systems. Source: https://openai.com/index/gpt-5-system-card-update-gpt-5-2