TOKYO - The Japanese government is considering requiring artificial intelligence-related businesses to take measures to ensure safety, with penalties for violations, a source familiar with the matter said Wednesday, as it seeks to address risks posed by advanced AI.
The proposal, presented at a closed-door working group meeting on AI-related legal systems, comes amid growing concerns about advanced AI models behaving unpredictably, as well as a raft of recent high-profile data leaks and unauthorized access incidents suspected to have involved the misuse of AI.
Under the envisaged approach, the government would be able to issue legally binding orders to businesses that do not cooperate with authorities following incidents involving AI. Noncompliance would be subject to penalties, which are expected to be limited to incidents involving advanced AI.
As AI-related businesses range from research firms and developers of foundation models to providers of services to end users, the government will examine which types of firms should be subject to the regulations.
The working group aims to present an overall outline of its review of the law, including possible revisions, by year-end.
Japan's AI law, enacted in May 2025, authorizes the government to investigate businesses and issue guidance in cases of malicious conduct but contains no penalty provisions, and its measures are nonbinding.
