Japan considers binding orders and penalties for AI firms over safety risks

The Japanese government is considering requiring AI businesses to take safety measures, with penalties for violations, Kyodo News reported, citing a source familiar with the matter. The proposal was presented at a closed-door meeting of a working group on AI legal systems.

Under the approach being considered, the government could issue legally binding orders to companies that do not cooperate with authorities after incidents involving AI, and noncompliance would be penalised. The penalties are expected to apply only to incidents involving advanced AI. The working group will also look at which kinds of companies, from foundation model developers to firms offering AI services to end users, should be covered.

Japan's AI law, passed in May 2025, lets the government investigate companies and issue guidance in cases of malicious conduct, but it has no penalties and its measures are not binding. The proposal follows concerns about advanced models behaving unpredictably and a string of data leaks and unauthorised access incidents suspected to involve misuse of AI. The working group plans to present an outline of its review, including possible changes to the law, by the end of the year.