Japan pushes to legislate AI safety obligations... New penalty for violation

In order to respond to risks such as cyber attacks caused by high-performance artificial intelligence (AI), the Japanese government is considering mandating AI business operators to ensure safety and establishing new penalties for violations.

According to the Mainichi Shimbun on the 12th, the Japanese government confirmed this policy at a working group meeting to discuss the AI ​​legal system held privately the previous day.

The Japanese government plans to prepare a full outline of the system reform, including legal revisions, by the end of the year.

This measure is intended to secure enforceable means to punish malicious and uncooperative business operators who do not respond to the government's request for correction even when a serious threat to security and safety occurs, such as when high-performance AI goes out of control or is abused for cyber terrorism.

The current AI Act stipulates that the government can investigate AI business operators when problems such as violations of citizens' rights occur, but there are no provisions for punishment.

The government is planning a plan to issue enforceable correction orders when necessary and apply penalties for violations.

However, it was decided to apply the penalty only to high-performance AI with high risk. We plan to respond to AI with general performance using existing laws and guidelines.

As the AI ​​business area is subdivided from basic model development to service provision, we plan to discuss the extent to which regulatory targets, including large U.S. information technology (IT) companies, will be included.

Meanwhile, at the closed meeting held the previous day, Antropic of the U.S., the developer of generative AI 'Claude', the Japanese subsidiary of OpenAI, the developer of 'ChatGPT', and Japanese telecommunications company NTT attended and delivered the status of their company's response and opinions.