AI Ethics Policy

1. Principles

   We believe AI is a technology that can contribute to human welfare and social progress, and we respect human dignity and human rights in how we develop and provide it. We pursue what AI makes possible, but never at the cost of safety or ethics.

2. Fairness and bias

   We work to ensure that our AI systems do not encourage unfair discrimination or prejudice on the basis of race, gender, age, nationality, religion, disability or similar grounds, at every stage of data collection, training and evaluation. We monitor continuously so that bias can be detected and corrected.

3. Transparency and accountability

   We make the decision-making of our AI systems transparent as far as is technically possible. Where AI output feeds into a significant decision, we build in a way to explain the basis for it as far as we can, and we hold ourselves accountable for it.

4. Safety and security

   We carry out proper risk assessment and security work when developing and operating AI systems. We test and improve continuously in order to minimise the risk of unexpected behaviour and of misuse.

5. Privacy

   We give the greatest possible respect to individual privacy in the data used to train and run our AI systems. Personal information is handled in line with our Privacy Policy and the applicable law, and we use only the minimum data necessary.

6. Human-centred AI

   The AI we build supports human decisions; it does not replace human judgement. As a matter of principle the final decision is always made by a person, and we design against over-reliance on AI.

7. Continuous review

   AI moves quickly and what society asks of it changes. We review and improve this policy regularly, and deepen our understanding of ethical questions through dialogue with outside experts and other stakeholders.

This English text is provided for convenience. The Japanese version is the authoritative text; if the two differ, the Japanese version prevails.