Japan AI Guidelines P-3.i: P-3) i. Consideration for bias in the configuration and data of AI systems and services

At provision, secure fairness of data and examine bias in referenced information and linked external services; periodically evaluate the model's inputs, outputs and basis of judgment to monitor for bias and, where needed, prompt the developer to re-evaluate bias in each technical element and decide on model improvement; and examine whether the systems, services and user interfaces receiving the model's outputs could embed bias that arbitrarily restricts business processes or the judgments of business users or non-business users.