Microsoft: China using AI to sway US elections

Microsoft researchers say they’ve discovered what appears to be a network of fake Chinese-run social media accounts set up to sway American voters with AI-generated images and other propaganda.

The researchers found the Chinese-linked “influencer” social media accounts, many found on Facebook and X (formally known as Twitter), are impersonating US voters with a level of sophistication not previously seen.

Calling the fake network “Chinese Communist Party (CCP) affiliated covert influence operations” the Microsoft research report published Thursday also said that American users are being targeted in greater numbers and with more direct engagement than in past campaigns.

Specifically, since 2022, the CCP-aligned accounts have been going after specific US candidates and posing as American voters to coincide with this November’s midterm elections, as well as next year’s US presidential election.

“The campaigns deploy thousands of inauthentic accounts across dozens of websites, spreading memes, videos, and messages in multiple languages,” the research shows.

In March, the Chinese-controlled accounts began to incorporate AI-generated images with the politically charged posts.

The Microsoft report provided several screenshots from the fake accounts on social media platforms Facebook and X, including this AI-generated image of the Statue of Liberty with a seven finger hand.

Facebook CCP  AI fake
AI-generated image posted by a CCP fake account on Facebook. Note: The Statue of Liberty’s hand holding the torch has more than five fingers. Image by Microsoft

When asked how Microsoft concluded the campaign was backed by Beijing, one researcher told Reuters news outlet that a "multifaceted attribution model" used for the analysis relied on "technical, behavioral, and contextual evidence."

Researchers said they were able to identify the CCP accounts because of commonly known characteristics used by Chinese social media trolls, for example, posting in Mandarin before switching to another language.

Microsoft said these type of accounts are often “operated by real people who employ fictitious or stolen identities to conceal the accounts’ affiliation with the CCP.”

These "real" influencers were also found to have direct ties to Chinese state media through factors such as employment, accepting travel invitations, or other monetary exchanges.

The fake accounts would list public locations within the US, post American political slogans, and share hashtags relating to domestic political issues, according to the report.

Twitter X CCP AI fake BLM
A Black Lives Matter graphic posted on X. Note: First uploaded by a CCP-affiliated automated account and then uploaded by an account impersonating a US conservative voter seven hours later. Image by Microsoft.

The new content is much more "eye-catching than the awkward visuals used in previous campaigns by Chinese nation-state actors, which relied on digital drawings, stock photo collages, and other manual graphic designs," the report said.

The researchers also found the AI-created visual content received more engagement from authentic American users even though it was obviously fake.

Microsoft said the Chinese political influencer campaign showed similar characteristics to the 912 Special Working Group, an “elite group” within China’s Ministry of Public Security (MPS).

The MPS made headlines in April after it was busted by the FBI for running at least half a dozen secret police stations in cities across the US – including New York City – to spread Chinese propaganda and intimidate Chinese-born US citizens.

The group is also known by the US Justice Department for operating a social media troll farm that created thousands of fake online personas and pushed CCP propaganda targeting pro-democracy activists, the research said.

In response to the report, a Chinese embassy spokesperson in Washington denied the accusations, saying that they were "full of prejudice and malicious speculation."

Separately, the Microsoft report also stated that China’s state-affiliated multilingual social media influencer initiatives have successfully engaged target audiences in at least 40 languages and grown its audience to over 103 million.

More from Cybernews:

Toyota factory stall caused by lack of disk space

VPN quagmire: latest industry gossip you need to know

Google, TikTok among EU's "gatekeeper" services

Using a password manager to secure Bitcoin

Spying made easier: China demands tech firms reveal weak spots

Subscribe to our newsletter

Leave a Reply

Your email address will not be published. Required fields are markedmarked