AI Misuse by Chinese Authorities Targeting Japan’s Leadership

A technology firm has reported that Chinese law enforcement attempted to misuse U.S. artificial intelligence to undermine Japanese political figures but faced a ban in the process.

In October 2025, OpenAI suspended a ChatGPT account tied to a Chinese official aiming to orchestrate a covert initiative against Japan’s Prime Minister, Sanae Takaichi. This revelation was shared by OpenAI in February 2026. Although the AI refused to participate, the user proceeded to outline what they termed a “cyber special operation.” OpenAI described these efforts as attempts to silence dissidents of the Chinese Communist Party and foreign leaders using methods like fake accounts, disinformation, and online harassment.

The campaign against Takaichi reportedly began after she publicly criticized human rights abuses in Inner Mongolia. The plan involved overwhelming social media with negative content, utilizing fake accounts masquerading as foreigners to influence Japanese politicians and paint Takaichi as a far-right extremist. Interestingly, it seems the campaign managed to continue without ChatGPT’s involvement, as similar AI-generated content and hashtags were found on platforms like X, YouTube, and Blogspot, according to OpenAI’s research.

Takaichi, notable as Japan’s first female prime minister and a strong critic of China, has drawn significant backlash from Beijing since assuming office in October. She remarked to a parliamentary committee that a Chinese attack on Taiwan could threaten Japan’s existence. In response, a Chinese diplomat in Osaka issued a now-deleted post on X, in which he allegedly made a violent threat against her.

OpenAI indicated that Chinese police were using ChatGPT to strategize and manage smear campaigns against their adversaries.

Additionally, officials reported that there are at least 300 individuals engaged in similar operations within one Chinese province, with estimates suggesting comparable numbers elsewhere. The claims included that these “special operations” were not limited to domestic platforms but extended to over 300 foreign sites. Reportedly, the operatives utilized Chinese AI models like DeepSeek and Qwen for monitoring, translation, and creating content.

These operatives also targeted prominent Chinese dissidents globally. Victims reportedly included activist Hui Bo and the human rights organization Safeguard Defenders. Techniques used ranged from filing mass fraudulent reports to creating fake accounts on Bluesky, even forging documents purportedly from U.S. courts.

The findings align with a 2023 lawsuit by the Justice Department that accused Chinese officials of orchestrating harassment and intimidation of dissidents abroad through fake social media personas and hacked livestreams.

Despite the extensive nature of this operation, it seems much of the content was ineffective. For instance, YouTube videos linked to the campaign received very few views, and out of an alleged 50,000 posts on Western platforms, fewer than 150 garnered meaningful engagement, as per OpenAI’s users. However, some dissidents did reportedly lose followers, halt their postings, or even have their accounts deleted due to the initiative.