OpenAI says it is working to protect what it calls “democratic AI” after uncovering and disrupting two influence operations that allegedly used ChatGPT to create content aimed at influencing public debate in the United States. The company said it banned “two clusters of ChatGPT accounts likely originating from China … after they used our models in support of apparent covert influence operations that promoted narratives in an attempt to manipulate a legitimate debate about American AI and wider tech policies”.“Our mission is to ensure that artificial general intelligence benefits all of humanity,” OpenAI said in a report released on June 10. By publishing its findings, OpenAI said it hopes to help governments, technology companies and the public identify and stop future influence operations.The company added that its mission includes “identifying and disrupting attempts by authoritarian regimes and their proxies to use AI systems to coerce critics, surveil communities or covertly interfere in democratic societies.”
Two campaigns originating from US targeted AI and technology debates: OpenAI
According to OpenAI, one of the campaigns focused on criticism of AI data centre projects in the US. The company said the operation generated social media comments and images claiming that AI data centres were driving up electricity costs for American families.OpenAI named this campaign “Data Center Bandwagon.” A second operation, called “Tech and Tariffs,” generated content criticizing US tariffs and technology policies.The company said prompts used by the operators specifically instructed ChatGPT to avoid mentioning Chinese President Xi Jinping and instead focus on US President Donald Trump. OpenAI also said the second campaign was connected to a network of likely inauthentic social media accounts that spread false claims that ChatGPT user data had been compromised.“These allegations were entirely false,” the company said.OpenAI did not disclose the exact number of accounts removed but said the campaigns involved hundreds of ChatGPT accounts that have now been banned.
OpenAI says influence operations had limited impact
OpenAI said it found no evidence that the campaigns significantly influenced public opinion. “The targeting of OpenAI and US data center buildouts is significant not because the operation appears to have shifted public opinion, but because it shows PRC-origin influence operators testing narratives against AI infrastructure,” the company said.The company added that the operators attempted to take advantage of existing concerns about energy prices and the impact of data centre development.
