OpenAI said it banned a ChatGPT account connected to an individual associated with Chinese law enforcement after that user sought help planning covert influence operations. According to OpenAI, the user attempted to craft a campaign targeting Japan’s prime minister, Sanae Takaichi, but the model refused to assist.
Operational planning and a disputed campaign
OpenAI reported the account asked the model in mid-October 2025 to help design a six-part operation aimed at discrediting Sanae Takaichi following her comments on human rights in Inner Mongolia. The proposed elements included posting and amplifying negative comments, fabricating complaints from foreign residents, leveraging cost-of-living grievances, accusing her of far-right links, stoking anger over U.S. tariffs, and promoting positive messages about conditions in Inner Mongolia.
Although ChatGPT declined to provide planning advice, OpenAI said the user later asked the model to polish a status report describing the campaign’s implementation. Open-source checks by OpenAI found activity that resembled the described operation: hashtags such as #右翼共生者 and a Japanese variant appeared on platforms including X, Pixiv, Blogspot and YouTube from late October 2025, and some memes and posts matched the themes in the report. OpenAI noted the posts showed low engagement, with the highest observed Pixiv view count at 108 and related YouTube videos recording single-digit views as of November 28, 2025.
Tactics, scale and earlier links
OpenAI said the user’s inputs also outlined a broad “cyber special operations” effort that reportedly employs hundreds of staff, thousands of fake accounts, locally deployed AI models such as DeepSeek and Qwen, and more than 100 tactics. The tactics listed included abusive reporting of dissident accounts, mass posting, forging documents, impersonating U.S. officials, doxxing, and targeting dissidents’ families.
OpenAI linked some observed activity to earlier campaigns. Meta attributed the Spamouflage operation to individuals associated with Chinese law enforcement in August 2023, and the U.S. Department of Justice made related accusations in 2023. Open-source reporting from October 2023 also documented prior campaigns that spread false obituaries and other harassment against activists such as Jie Lijian.
OpenAI said the evidence suggests a sustained, resource-intensive pattern of covert influence activity targeting dissidents, critics of the CCP and foreign officials, but noted that much of the identified online activity achieved limited visible impact.
Original source: OpenAI News