China Is Using AI to Influence Elections, Microsoft Warns

The Chinese government continues to refine its use of artificial intelligence tools to conduct information operations aimed at sowing disruption and influencing global elections, security researchers warned.

See Also: Generative AI Survey Result Analysis: Google Cloud

Beijing seems intent on influencing elections in the United States as well as spreading discord, especially in Taiwan, Japan and South Korea, says a report from the Microsoft Threat Analysis Center.

With more than 1 billion people across more than 50 countries – including the U.S., United Kingdom and India – poised to hold elections this year, security experts are sounding warnings about interference efforts. U.S. intelligence says the four nation-states that pose the biggest cyber risk to American national security – Russia, China, Iran and North Korea – also pose the greatest threat to the integrity, or perceived integrity, of American elections (see: As Elections Loom, So Do Adversaries’ Influence Operations).

China-backed advanced persistent threat groups are testing and refining fresh tactics for influencing global elections, including via the use of AI-generated deepfakes, albeit with apparently low levels of success so far, Microsoft said.

Malicious experimentation is rife. Taiwanese presidential elections in January for the first time ever featured “a nation-state actor using AI content in attempts to influence a foreign election,” said Clint Watts, head of the Microsoft Threat Analysis Center, in a blog post.

A Beijing-affiliated group tracked as Storm-1376 – also known as Spamouflage and Dragonbridge – posted fake audio, potentially generated using AI tools, of former presidential contender Terry Gou endorsing another candidate on Election Day. “Gou had made no such statement,” Watts said. “YouTube quickly removed this content before it reached a wider audience.”

Other tactics attributed to Beijing during the election include the use of memes and further audio content and videos, including AI-generated news anchors. That followed what appeared to be a dry run last summer, when Microsoft reported seeing a surge in Chinese-affiliated groups running campaigns that targeted U.S. political leaders, seeking to cause division and disruption. English-language accounts it tied to the Chinese Communist Party issued a variety of false claims, including that the U.S. government triggered the August 2023 wildfires in Hawaii while testing a new military “weather weapon” and caused the November 2023 train derailment in Kentucky that led to molten sulphur catching fire.

Since then, multiple social media accounts tied to the CCP continue to “nearly exclusively post about divisive U.S. domestic issues such as global warming, U.S. border policies, drug use, immigration and racial tensions” and make free use of “original videos, memes, and infographics as well as recycled content from other high-profile political accounts,” Watts said.

While China appears to have been testing the use of fake AI anchors since at least February 2023, “the volume of its content featuring these anchors has increased in recent months,” Microsoft said. For the Taiwanese elections, the researchers said, the AI-hosted segments often included footage of the country’s officials and appeared to have been generated by Storm-1376 using off-the-shelf tools such as ByteDance’s CapCut tool, which enables users to create AI-generated videos using a single prompt.

China’s election targeting crosses over at least in part with its espionage operations focusing on targets of military and economic interest, Microsoft said. Broadly speaking, it sees Beijing focusing on three espionage areas: America’s industrial base, the South China Sea region – especially Taiwan and Malaysia, and “nearly every South Pacific Island country,” including Papua New Guinea.

Microsoft’s alert comes on the heels of the U.S. Cybersecurity and Infrastructure Security Agency warning of a rise in election interference efforts as part of “broader efforts to undermine U.S. global standing, sow discord inside the United States, and influence U.S. voters and decision-making.”

Where possible, China appears to be increasingly keen to test the impact of AI-generated content as part of its influence operations.

“With major elections taking place around the world this year, particularly in India, South Korea and the United States, we assess that China will, at a minimum, create and amplify AI-generated content to benefit its interests,” Watts said. “Despite the chances of such content in affecting election results remaining low, China’s increasing experimentation in augmenting memes, videos, and audio will likely continue – and may prove more effective down the line.”