OpenAI report links ChatGPT use to alleged Chinese campaign targeting dissidents abroad

A Chinese influence operation aimed at silencing critics overseas was exposed in an unexpected way — through what OpenAI says was a Chinese law enforcement official’s habit of using ChatGPT as a running log of the campaign.

According to a new report from the company, the user treated the chatbot like a private diary, recording tactics, targets and progress in what investigators describe as a coordinated effort to intimidate Chinese dissidents living outside the country. OpenAI says it shut down the account after identifying the activity.

The entries outline a campaign that went beyond online harassment. In one case, operatives allegedly posed as US immigration officials and warned a dissident in the United States that their public statements had violated the law. In another, the user described producing forged documents that appeared to come from a US county court in an attempt to pressure social media platforms into removing a critic’s account.

OpenAI said its analysts were able to connect those descriptions to real-world activity. The user wrote about fabricating an obituary and gravestone images to create the impression that a dissident had died; false rumours about that person’s death later circulated online in 2023. The operation itself, the report says, involved hundreds of participants and thousands of fake accounts across multiple platforms.

“This is what Chinese modern transnational repression looks like,” Ben Nimmo, a principal investigator at OpenAI, said ahead of the report’s release. “It’s not just digital. It’s not just about trolling. It’s industrialized. It’s about trying to hit critics of the CCP with everything, everywhere, all at once.”

The material generated through ChatGPT was not the primary content of the campaign. OpenAI said the tool functioned mainly as an internal record-keeping system, while posts and images were created using other technologies and then distributed across social media and websites.

In one instance, the user asked the chatbot to design a plan to discredit Japan’s incoming prime minister, Sanae Takaichi, by amplifying anger over US tariffs on Japanese goods. ChatGPT declined to provide an answer, the company said. Later, as Takaichi took office, online hashtags attacking her and referencing US tariffs appeared on a forum for Japanese graphic artists.

OpenAI’s findings arrive as Washington and Beijing compete for technological advantage in artificial intelligence, a rivalry that extends from military applications to information campaigns. Michael Horowitz, a former Pentagon official now at the University of Pennsylvania, said the case shows how AI tools can be folded into existing state systems. “US-China AI competition is continuing to intensify,” he said, adding that the technology is being used not only at the cutting edge but also in the “day-to-day of their surveillance and information apparatus.”

The report also lands in the middle of a separate dispute between the US Defense Department and AI company Anthropic over access to its model, a sign that the question of how these systems are controlled and deployed is becoming as important as who builds them.