Bogus AI Images Used By Pro-China Accounts To Inflame U.S. Opinion

(Bloomberg) -- Chinese groups are using social media and artificial intelligence-generated images to stoke controversy in the US during an election year and gain insight into how Americans view divisive issues, according to Microsoft Corp. findings.

Social media accounts linked to China have pushed conspiratorial content, such as blaming the US government for summer 2023 wildfires on the Hawaiian island of Maui, the company said in a report published Friday. Within days of the wildfire’s kindling, the report says, one Chinese group pushed out posts in 31 languages, alleging the US government had deliberately set the blaze.  Those posts included AI-generated images of burning coastal roads.

An official probe into the cause of the fire that destroyed the seaside Hawaiian town of Lahaina is ongoing, but investigators have focused on whether power lines owned by Hawaiian Electric Industries Inc. sparked the flames.

The fabricated images signal that Chinese government-affiliated groups are adopting new methods as the country pursues longstanding geopolitical priorities through a continued campaign of disinformation and cyberattacks, Microsoft said. It’s unclear, however, whether artificial intelligence has meaningfully enhanced the reach or power of those efforts.  

Related:Top Tips for Cybersecurity Tabletop Exercises and Simulations

Since last fall, Microsoft  said in the report, there’s been a “small yet steady increase” in social media accounts linked to China pushing inflammatory talking points. The Redmond, Washington-based company also found Chinese influence campaigns focused on Taiwan’s election and more broadly on deepening rifts in the Asia-Pacific region.

On election day in Taiwan, a Chinese-associated propaganda group “likely” used an AI-generated audio recording in which Terry Gou, owner of Foxconn Technology Group and a former presidential candidate, appeared to endorse another candidate, according to the report. The report says the Chinese group behind the Maui fires posts also pushed out messages to cast doubt on the safety of Japan’s disposal of radioactive wastewater, making specific efforts to reach people in South Korea with a campaign that “bears the hallmarks of AI generation.”

Microsoft “assess[ed] with moderate confidence” that the accounts posting this material are run by the Chinese government. A company spokesperson said it has concluded that the Chinese government commands some of the groups identified in the report , while others are located in China and use targeting or techniques that align with a state group.

The Chinese Embassy didn’t return a request for comment. The government typically denies any connection to such activity. 

Misinformation researchers last year said they detected AI-generated images of news anchors promoting Chinese Communist Party interests.

The use of AI to produce eye-catching images and doctor video hasn’t necessarily proven persuasive. Microsoft’s report finds that these efforts have gained varying degrees of traction and that “the impact of such content in swaying audiences remains low.”

These efforts have dovetailed with warnings from the US government that Chinese hacking groups have been infiltrating critical infrastructure, including communications, energy, transportation and water systems. Microsoft said one such group has also targeted government agencies and companies related to national defense, although it's unclear whether these attacks have been successful. 

Microsoft was the subject of a recent US government report that criticized the company’s response to a suspected Chinese cyber-espionage campaign.