Japan's government AI must not serve as tool for 'digital militarism'

Editor's note : Xiang Haoyu is a specially appointed research fellow at the Department for Asia-Pacific Studies, China Institute of International Studies.

The article reflects the author's opinions and not necessarily the views of CGTN.

In recent years, with the rapid development of AI technology, generative AI has become an important tool for governments worldwide to promote administrative digitization.

Reportedly, Japan's Digital Agency is deploying a government-dedicated generative AI system named "Gennai" on a large scale, which is expected to cover approximately 180,000 central government civil servants.

On the surface, this appears to be a technological innovation aimed at improving administrative efficiency.

However, a closer analysis of its underlying model dependencies, core data sources, and application scenarios reveals an alarming trend.

If a government AI system relies on selectively curated historical data and foreign-developed foundation models, it may not simply automate existing administrative practices; it may also reproduce and amplify the political assumptions embedded in those systems.

Without adequate safeguards, "Gennai" could become vulnerable to political manipulation and, in the process, contribute to historical revisionism and what might be called "digital militarism." The 'truncated' historical database The objectivity of AI-generated content depends heavily on the data used to train, retrieve and ground the system.

The Japanese government has fed approximately 18 million pages of government data into the "Gennai" system, with one of the core corpora being data from Japan's Official Gazette (Kanpō) spanning from May 1947 to the present.

The choice of this specific starting point carries heavy political significance in terms of historical narrative.

May 1947 marks the implementation of Japan's post-war Peace Constitution.

By anchoring the data's starting point precisely at this moment, the historical database of the "Gennai" system is physically and logically "truncated." This implies that key historical archives – such as the Imperial Rescript on the Termination of the War regarding Japan's 1945 surrender and vast records of aggression committed by Japanese militarism against Asian nations during World War II – are very likely systematically excluded from the AI's core knowledge graph.

This data gap directly causes "Gennai" to lack the fundamental framework of post-war defeat characterization and historical reflection when addressing modern history, state responsibility and territorial disputes.