Chinese Operatives Using AI for Propaganda
On Thursday, OpenAI released a threat report indicating that Chinese operatives are leveraging ChatGPT to advance propaganda aimed at American political conversations.
The report suggested that these actors are using the model to engage with contentious American issues, mock former President Donald Trump’s tariffs, and disseminate politically charged content on platforms like TikTok, X, Reddit, and Facebook. OpenAI determined that four out of ten malicious accounts likely have ties to China.
Ben Nimmo, an investigator at OpenAI, commented on the diverse tactics observed from China, noting, “They employ various strategies.” He also highlighted how these operations target a range of countries and subjects, mixing elements of influence operations, social engineering, and surveillance across multiple online platforms.
There is a particular effort dubbed “sponsor’s dissatisfaction,” which involved generating tweets that criticized opposition figures. Additionally, articles were translated from Chinese and published in Spanish media outlets.
The operations not only utilized ChatGPT for creating phishing emails but also translated propaganda from Latin American sources. One campaign, referred to internally as “Sneer Review,” was responsible for producing numerous political comments across platforms in different languages.
Moreover, some efforts sought to impersonate both sides of the American cultural debate, intensifying discord on sensitive topics. One tactic involved establishing fake accounts that gave the content an appearance of grassroots support.
Chinese operatives have also targeted a niche Taiwanese video game named “Reversed Front,” generating negative comments in Chinese and even crafting a fabricated long-form article claiming it sparked massive public outcry.
OpenAI’s report noted that threat actors mimic internal documents, with one instance being a detailed essay that emulated public safety communications, discussing how members of Chinese safety organizations promote self-discipline while adhering to Xi Jinping’s legal principles.
The Beijing Embassy in Washington did not respond to inquiries for comments.
Beyond China, OpenAI has also disrupted smaller influence campaigns tied to marketing firms in Russia, Iran, North Korea, and the Philippines. Most of these countries’ embassies did not reply to requests for comments, with North Korea being an exception due to its lack of a U.S. embassy. Notably, OpenAI has not shared details on failed campaigns that did not engage real users before being dismantled.
The report stated, “Although there was activity across multiple platforms, engagement was minimal and the content rarely shared.”
OpenAI did not respond to inquiries for further comments.
