September 19, 2025
Tokyo – The government will support the development of domestic artificial intelligence models based on Japanese data and technology to avoid reliance on AI from the US and China, according to sources.
The US and China are making progress in developing generator AIs that can create text and other content, but there are concerns that rely on these overseas models, which may cause misinformation about how it is spreading overseas and Japan is widely used. This is considered a problem from a security perspective.
Resources are provided to Japanese companies for the development of learning data and other features to support the creation of what the government wants to become a highly reliable domestic AI model.
The National Institute of Information and Communications Technology (NICT) is under the jurisdiction of the Ministry of Interior and Communications and provides Japanese data collected over the past 20 years. Nict and Preferred Networks, Inc, is a Tokyo-based AI development company. will collaborate on developing AI models that will produce highly reliable answers to Japanese culture, customs, systems and other fields.
The government expects information technology company Sakura Internet Inc. to offer new domestic AI models through its data center in Japan.
The Ministry of Communications and the Ministry of Economics, Trade and Industry plan to provide financial and other support for other issues, including providing the infrastructure necessary to develop the model, recruiting talented engineers with exceptional capabilities, and maintaining high-quality Japanese data. It is expected that central governments, local governments and businesses will use the Japanese AI model.
Japanese companies and organizations handle all aspects of data centre operations as they aim to provide data used for AI learning, to develop “large language models” that will become the brains of models, and governments aim to create and provide generative AI models entirely within the country.
In Japan, the use of US and Chinese AI models is growing. However, the learning data and other factors in the development process for these overseas models are unknown, raising concerns among Japanese government officials. “These models are fearful of learning data that is unacceptable from a Japanese perspective,” a government source said.
The overseas model mainly learns data from English sources. This states that some observers do not understand Japanese culture, history and other areas. Another problem is that the principles and opinions of the countries that developed these AI models can influence the answers they generate.
In one case, Ithunori Onodera, chairman of the liberal Democratic Council of Policy Research, asked China's Deepseek-generated AI model whether the Senkaku Islands in Okinawa are Japanese territory. Deepseek replied that the islands are “inherent territory of China.”
Believing that reliance on overseas AI models presented risks, the government decided to support the development of AI models that provide highly reliable answers to Japanese culture, customs, history and other areas.
AI development requires a huge amount of data, computers and excellent HR with expertise. However, Japan is delaying the US and Chinese local leaders and China in this regard.
According to the Artificial Analytics Intelligence Index, an international composite metric for assessing generative AI capabilities, US and Chinese AI services such as Openai and Alibaba dominate the sector. More and more Japanese companies use these US and Chinese AI models as bases, raising concerns about their reliance on services from overseas.
