In line with the characteristics and uses of the AI system or service, secure the accuracy and, where needed, the recency of the data used for training (appropriateness of the data); within a reasonable range, properly ensure transparency of training data, compliance with legal frameworks and updating of AI models; and take care that illegal content such as infringing reproductions (pirated copies are the rights holders' concern cited) is not included in training data. Providers and business users who fine-tune or retrain are also expected to ensure safety as developers do.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.