Japan Eyes 10 Principles for Using AI in Business; Draft Guidelines Urge Transparency, Fairness

Reuters file photo
ChatGPT’s logo is seen in this illustration taken in February.

The government has laid out 10 principles related to artificial intelligence, including “fairness” and “transparency,” in draft guidelines for companies and organizations involved with AI, according to sources.

The draft guidelines were unofficially presented Tuesday at a meeting of the AI strategy council, an expert panel. They are expected to be finalized by the end of the year, following discussions by the council.

At the end of October, the Group of Seven industrialized nations agreed on international guiding principles for AI and a voluntary code of conduct for AI developers. Japan’s guidelines are intended to put the code into practice domestically.

According to the draft, the guidelines target all those who use AI for business, including public organizations such as ministries and agencies. Users who do not directly utilize AI in their businesses, as well as data providers, are not covered by the draft guidelines.

The 10 principles begin with the principle of being “focused on human beings,” which requires AI-related operators to “respect the dignity of individuals and pay careful attention to potential harm.” It also calls on them to “not develop or provide AI for the purpose of wrongfully manipulating human decisions or emotions.”

In terms of countering disinformation, the guidelines state that these parties should “take necessary measures” in recognition of the risk of causing social disruption.

The “fairness” principle noted “the possibility that AI learning data may contain unacceptable biases,” and emphasized the “factoring human judgment into its use” to ensure that AI does not promote discrimination.

With respect to data learning by AI, the “transparency” principle calls for the operators to provide information about data collection methods, and the “accountability” principle states that the source of data should be “traceable to the extent technically feasible.”

However, the guidelines did not stipulate specific issues to be addressed, such as concerns about copyright infringement caused by AI learning. Ensuring the effectiveness of the AI guidelines may be an issue to be tackled ahead of their formulation by the end of the year, observers said.

In addition to the 10 common principles, the draft included “documentation of the development process in a form that can be verified by a third party” as a point of caution for AI developers.

The government is believed to be envisioning a certification system by a third party, according to the observers.

The guidelines also include notes for operators that provide services that incorporate AI, as well as businesses that use AI services.