OpenAI Report Exposes Chinese Use of AI for Transnational Influence Operations and Repression

OpenAI disclosed on February 25, 2026, that accounts linked to Chinese law enforcement had used ChatGPT as a work log for “special cyber operations” and to polish reports on cross-border opinion manipulation and the suppression of dissidents. The operation involved hundreds of personnel, thousands of fake accounts and more than 100 tactics. It also targeted pro-Taiwan commentary, underscoring Taiwan’s need to guard against AI amplifying existing information warfare.

In the latest development, OpenAI blocked two more groups of ChatGPT accounts suspected of originating in China on June 10, 2026. One promoted claims that AI data centers were driving up U.S. household electricity prices, while the other framed technology competition and Trump’s tariffs as sources of public anger. Both generated social-media posts and images at scale. OpenAI said it had found no evidence that the content spread beyond the operators’ own network of accounts or materially influenced public opinion.

OpenAI has issued threat reports over the past two years tracking the misuse of generative AI in scams and influence operations. These threats do not rely on a single model. Instead, perpetrators combine ChatGPT, websites, social-media accounts and other AI tools into workflows that make existing scams easier to translate, automate and distribute, raising concerns for platform governance and cross-industry information sharing.

OpenAI released its latest report and banned accounts linked to the activity on February 25, 2026. A person connected to China’s law-enforcement system asked the model to plan a covert influence operation involving dozens of tactics, with the model refusing at least once. A Cambodian network combined human prompts with automated chatbots to target Indonesian men and was estimated to deceive hundreds of people each month. Payment demands increased incrementally, but the report did not disclose the amount lost.