China conducts “influence campaigns” using content generated by artificial intelligence (AI), says a report from the American computer company Microsoft, published this Thursday, April 4. While elections will take place in India, South Korea and the United States in 2024, “China will, at a minimum, create and amplify content generated by AI to serve its interests”, specifies the Microsoft Threat Analysis Center blog post, who presents the report. “While such content is unlikely to influence election results, China’s growing experiences […] could prove more effective in the future,” warns the document.
According to the filing, online actors linked to the Chinese government are increasingly using AI to gather information on topics that divide opinion in different countries. For example, they use fake social media accounts to “ask contentious questions on controversial US national issues in order to better understand the key issues that divide American voters. This could include gathering intelligence and details on main demographic characteristics of voters before the US presidential election”, analyzes the Threat Analysis Center.
“These tactics have been used in campaigns aimed at stoking divisions in the United States and exacerbating dissension in the Asia-Pacific region, including Taiwan, Japan and South Korea,” the report details.
Disinformation and threats to elections
More concretely, the report explains that these accounts falsely claimed that the Maui (Hawaii) fires of August 2023 “were deliberately started by the US government to test a military-grade weather weapon”, using photos of Giant fires generated by AI. Posts of this type then ask their subscribers to react, which would allow them to know their profile and what they think about the subject. Information that could be used “to improve intelligence collection on key demographic groups who vote in the United States,” according to Microsoft.
The Taiwanese presidential election of January 2024 was also marked by “a sharp increase in the use of content generated by AI” during operations carried out by actors affiliated with the Chinese Communist Party, specifies Microsoft: “This is the first time Microsoft Threat Intelligence has observed a state actor using AI-generated content to attempt to influence a foreign election.” A fake AI-generated audio tape was posted on YouTube on election day, in which election candidate Terry Gou, who withdrew from the race in November 2023, supported another candidate for president – a statement that did not exist.
Already in September 2023, the Analysis Center reported that fake accounts had begun to impersonate American voters in an attempt to influence the 2022 US midterm elections.
Limited impact, for now
As people in India, South Korea and the United States head to the polls this year, the report finds it “likely that cyber actors with Chinese influence” are working “to target these elections.” However, he recalls that “the impact of this type of content on the public remains low”, but that the increase in experiments “could prove effective in the long term”.
The report comes as a White House-appointed review panel “said a ‘cascade of errors’ by Microsoft allowed state-backed Chinese cyber operators to break in in the email accounts of senior American officials”, recalls the British newspaper The Guardian. In March, the US and UK governments accused China-backed hackers of carrying out a years-long hacking campaign targeting politicians, journalists and businesses.