OpenAI has revealed it recently shut down ten covert influence operations that were using its artificial intelligence tools, including ChatGPT, to spread propaganda and conduct online surveillance. Four of these campaigns have been linked to actors associated with the Chinese state.
These operations exploited generative AI to produce and distribute propaganda across major social media platforms such as TikTok, Facebook, Reddit, and X. The campaigns used AI to generate social media posts, create fake engagement by replying to their own content, and even draft internal performance reviews to document their activities.
One notable Chinese operation, nicknamed “Sneer Review” by OpenAI researchers, created short comments in English, Chinese, and Urdu. These posts targeted topics ranging from U.S. politics to criticism of a Taiwanese video game that portrays the defeat of the Chinese Communist Party. The group aimed to simulate authentic online engagement by posting both original content and replies.
Another Chinese-linked network posed as journalists and geopolitical commentators. They used ChatGPT to create fake account biographies, translate communications between Chinese and English, and analyze correspondence addressed to a U.S. Senator about a federal nomination. This same network also used AI to create marketing materials advertising fake social media campaigns and intelligence recruitment services. OpenAI previously reported that similar groups had used AI to produce anti-American articles in Spanish, which were then published in Latin American news outlets.
Despite the advanced technology involved, OpenAI found that these influence operations had little real-world impact. Most of their content failed to attract genuine engagement, and in many cases, the only responses came from users who recognized the posts as fake. This finding aligns with similar reports from Meta, which noted that even with AI, bad actors still face the same challenges in convincing real people to interact with their propaganda.
OpenAI’s latest disclosures highlight the growing use of generative AI in global information warfare, but also suggest that technology alone does not guarantee successful manipulation of public opinion. As Ben Nimmo, principal investigator on OpenAI’s intelligence team, noted: “These operations may be using new technology, but they’re still struggling with the old problem of how to get people to fall for it”.
One Chinese operation, dubbed “Sneer Review” by OpenAI researchers, generated short comments in English, Chinese, and Urdu targeting diverse subjects from U.S. politics to criticism of a Taiwanese video game where players defeat the Chinese Communist Party. The campaign created both original posts and replies to simulate authentic engagement, according to the company’s threat report.
A separate Chinese operation involved actors posing as journalists and geopolitical commentators who used ChatGPT to create account biographies, translate communications between Chinese and English, and analyze correspondence addressed to a U.S. Senator regarding a federal nomination. … This follows OpenAI’s February report documenting Chinese groups using AI to create Spanish-language anti-American content published in Latin American news outlets, marking what the company called the first known instance of Chinese influence operations translating long-form articles into Spanish for regional distribution.
Despite the sophisticated use of AI technology, OpenAI reported the operations failed to achieve substantial reach or authentic engagement. In some cases, the primary interaction came from users identifying the content as fabricated. … “These operations may be using new technology, but they’re still struggling with the old problem of how to get people to fall for it,” Nimmo said.
