Japan is drawing up plans to penalize artificial intelligence companies that disregard government orders, a measure aimed primarily at U.S. developers of advanced AI whose systems Japan relies on but cannot compel to share information. Officials intend to publish a full reform blueprint before the end of 2026.

Why Japan struggles to make American AI companies cooperate

Japan's AI Promotion Act, which allows the government to investigate and issue guidance when AI misuse harms someone's rights, took full effect in September 2025. The law, however, contains no punishment for a company that simply refuses to cooperate, and Japan is now weighing whether to write penalties into the statute.

The regulatory gap is more pronounced with foreign firms. In a draft proposal, a group within Japan's ruling Liberal Democratic Party, led by former digital minister Masaaki Taira, warned that while other governments can obtain information from AI companies, Japan may be unable to do the same. The draft noted that the European Union already fines AI companies that refuse to hand over information — the kind of enforcement lever Japan's current statute lacks.

Japan's concern centers on advanced AI that it neither builds nor controls. The government wants AI companies to assume safety obligations and is considering penalties tied to the risk of advanced AI going out of control or being used in cyberattacks. A private working group has confirmed the plan, and officials also want the authority to act against companies that decline to assist after an incident occurs.

Japan's Financial Services Agency reportedly formed a 36-member group of government and industry representatives in May 2026 to study cyber threats linked to Anthropic's Mythos model, with the Japanese branches of Anthropic and OpenAI taking part. Japan's updated AI Basic Plan, approved in July 2026, also named high-performance models such as Claude Mythos as a concern. Separately, the U.S. Commerce Department limited access to Anthropic's Mythos 5 and Fable 5 earlier in 2026, then restored it.

The government reportedly began reviewing the AI Promotion Act at an expert panel meeting on October 26. Digital minister Toshiharu Furukawa told the panel that Japan needs closer work with partner countries on AI safety. A group of academics held its first meeting two days later, and the panel hopes to settle on a direction by the end of the year.

It is not yet clear that penalties will survive the process. Some participants have suggested that Japan proceed with caution, because other countries are not all considering tighter AI rules.

Will Japan's penalties be big enough to matter?

The scale of any fine would shape its effect. South Korea's AI Basic Act, which took effect earlier this year in January, lets regulators fine companies up to 30 million won (about $21,000) for refusing inspections or ignoring corrective orders. For a large U.S. AI company, a fine that small is negligible and unlikely to change behavior.

Japan initially took the opposite approach. Passed in May 2025 as part of an effort to make the country the “most AI-friendly country,” its first AI law, the AI Promotion Act, deliberately left out penalties so that cautious firms would not be scared away from using AI. The law was designed to work through social pressure and voluntary compliance, and analysts assumed that a company like OpenAI would have to follow local rules if it wanted access to Japan's market.

By the end of the year, it will be clear how large Japan's penalties would be — or whether there will be any at all. The full blueprint due before the end of 2026 will also show whether regulators would gain the authority officials have sought to act against companies that decline to assist after an incident.