OpenAI says ChatGPT refuses Chinese influence operation

OpenAI reported that its ChatGPT service refused to assist an individual associated with Chinese law enforcement in planning an online campaign to discredit Sanae Takaichi, whom OpenAI described as Japan’s first female prime minister.

The company said the user requested help to amplify negative comments about her and portray her as far-right amid what OpenAI called a broader, large-scale, resource-intensive, and sustained misinformation effort.

OpenAI also flagged other activities, including romance scams targeting Indonesians, a content farm linked to Russia, and accounts deemed likely to have originated in China seeking information from US officials.

OpenAI says the plan surfaced in mid-October, in the days leading up to her election, when she criticized human rights in Inner Mongolia.

The prime minister later drew a rebuke from China over Taiwan remarks, a diplomatic spat affecting tourism and trade.

Chinese foreign ministry spokeswoman Mao Ning said she was not familiar with the accusations and saw no basis for them.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Chinese operatives used ChatGPT as a ‘diary’ for their global intimidation campaign

A Chinese law enforcement official used one ChatGPT account as a private journal to track a wide, multi-platform effort aimed at critics overseas 1.
The activity went beyond misinformation. It included posing as U.S. immigration officials to threaten dissidents, plus fabricating U.S. court papers to push platforms to remove social media accounts 1.
Targeting of Sanae Takaichi appeared to continue after ChatGPT refused to assist. Later prompts suggested the effort moved ahead. OpenAI also tied related activity to other tools, including Chinese AI models like DeepSeek 2.

Early-stage influence operations can still foreshadow AI-enabled authoritarianism risks

OpenAI researchers said these covert efforts, even with AI, were mostly disrupted early and generally failed to reach large audiences 3.
Even so, the case matches a long-term concern raised by AI leaders. Autocracies could use powerful AI for mass surveillance and influence campaigns 4.
That concern drives work on AI safety. Labs are building in constitutions, sets of rules that constrain model behavior. Some national security researchers have argued for adding respect for sovereignty directly into AI systems as a restraint strategy 5.