Because publicly sourced training data may have been poisoned and the poisoning is usually found, if at all, only after training, developers record how they obtained public training data, its origin, and the way the model uses it. The record should at least hold the source (such as the URL of a scraped page) and the date and time the data was collected, so that a reported poisoning attack can be checked against the organisation's own datasets.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.