Thursday, February 22

The pro-China YouTube network used artificial intelligence to defame the United States, a report finds

In a somewhat stilted tone and somewhat awkward grammar, the American-accented voice on YouTube last month ridiculed Washington’s handling of the war between Israel and Hamas, arguing that the United States was not in capable of “playing its role as a mediator like China” and “now finds itself in a position of significant isolation”.

The 10-minute post was one of more than 4,500 videos in an unusually large network of YouTube channels spreading pro-China and anti-American narratives, according to a report this week by the Australian Strategic Policy Institute, a security-focused think tank.

Some of the videos used artificially generated avatars or voiceovers, making the campaign the institute’s first known influencer operation to pair AI voices with video essays.

The goal of the campaign, according to the report, was clear: to influence world opinion in favor of China and against the United States. The videos promoted the narrative that Chinese technology was superior to American technology, that the United States was doomed to economic collapse, and that China and Russia were responsible geopolitical actors. Some clips flattered Chinese companies like Huawei and denigrated American companies like Apple.

The report notes that content from at least 30 of the network’s channels has garnered nearly 120 million views and 730,000 subscribers since last year, along with occasional advertisements from Western companies.

Some videos featured titles and writing that appeared to be direct translations of common Chinese phrases and Chinese company names, the report said. Others mentioned information that could be traced back to news stories produced and disseminated primarily in mainland China.

Misinformation – such as the false claim that some Southeast Asian nations had adopted the Chinese yuan as their currency – was common. The videos were often able to react quickly to current events. Jacinta Keast, an analyst at the Australian institute, wrote that the coordinated campaign could be “one of the most successful China-related influence operations ever seen on social media.”

YouTube said in a statement that its teams are working around the clock to protect its community, adding that “we have invested heavily in robust systems to proactively detect coordinated influence operations.” The company said it welcomed the research efforts and had shut down many of the channels mentioned in the report for violating platform policies.

Efforts to promote pro-China messages have proliferated in recent years, but have featured largely low-quality content that has attracted limited engagement or failed to sustain significant audiences, Keast said.

“This campaign actually leverages artificial intelligence, which gives it the ability to create persuasive threat content at scale at a very limited cost compared to previous campaigns we’ve seen,” he said.

Several other recent reports suggest that China has become more aggressive in pressuring disparaging U.S. propaganda. Historically, its influence operations have focused on defending the Communist Party government and its policies on issues such as the persecution of Uighurs or the fate of Taiwan.

China began targeting the United States more directly during mass pro-democracy protests in Hong Kong in 2019 and continuing through the Covid-19 pandemic, echoing long-standing Russian efforts to discredit American leadership and influence at home and abroad.

Over the summer, researchers at Microsoft and other companies uncovered evidence of inauthentic accounts used by China to falsely accuse the United States of using energy weapons to start deadly wildfires in Hawaii in August.

In a September report, the State Department accused China of using “deceptive and coercive methods” to shape the global information environment, including creating fake social media accounts and even fake news organizations. Other research suggests that China has been actively spreading disinformation in Taiwan that the United States will eventually betray the island nation.

Meta announced last month that it had removed 4,789 Facebook accounts from China posing as Americans to discuss political issues, warning that the campaign appeared to be laying the groundwork for interference in the 2024 presidential election. It is the fifth network with ties to China taken over by Meta this year, more than any other country.

The advent of artificial technology seems to have aroused particular interest on the part of Beijing. The Australian institute’s Keast said peddlers of misinformation are increasingly using easily accessible video editing and artificial intelligence programs to create large volumes of convincing content.

He said the network of pro-China YouTube channels most likely inserted English-language scripts into text-to-video conversion software readily available online or into other programs that require no technical skills and can produce clips in minutes. Such programs often allow users to select AI-generated voice narration and customize the gender, accent and tone of the voice.

Some of the rumors used in the pro-China network were clearly synthetic. Keast noted that the audio lacked natural pauses and included mispronunciations and occasional notes of electronic interference. Occasionally, multiple channels in the network used the same voice. (One group of videos, however, tried to trick viewers into believing a real person was speaking, embedding audio like “I’m your host, Steffan.”)

In 39 videos, Keast found at least 10 artificially generated avatars advertised by a British artificial intelligence company. You wrote that you also discovered what may be the first example of a digital avatar influence operation created by a Chinese company: a woman in a red dress named Yanni.

According to the report, the reach of the pro-China network is likely even broader. Similar channels appeared to target Indonesians and the French. Three separate channels posted chip manufacturing videos that used similar thumbnail images and the same title translated into English, French, and Spanish.