AI TRENDS | Japan Considers Penalties for AI Firms Over Safety Risks

Japan is considering requiring AI-related companies to implement security measures and imposing penalties for violations to address risks from advanced AI, people familiar with the matter said Wednesday. According to Sina Finance, the proposal was raised at a closed-door working group meeting on the AI legal framework, amid growing concerns that advanced AI models could behave unpredictably and recent high-profile data leaks and unauthorized access incidents suspected of being linked to AI misuse. Under the proposal, if an AI-related security incident occurs and a company refuses to cooperate with regulators’ investigation, the government could issue a legally binding order to take corrective action. Companies that fail to comply could face penalties, which are expected to apply only to incidents involving advanced AI. The entities covered could include developers of foundational AI models, research companies and providers of AI services for end users; the Japanese government plans to define the scope of regulated companies further. The working group plans to produce an overall outline of the legal review, including possible amendments, by the end of the year. Japan’s current AI law took effect in May 2025. It allows the government to investigate malicious conduct and provide administrative guidance, but does not impose penalties, and the measures are not legally binding.