China Using AI to Create Anti-American Memes Capitalizing on Israel-Palestine, Researchers Find

A campaign linked to the Chinese Communist Party is creating and spreading AI-generated memes designed to stoke anti-American sentiment by capitalizing on President Biden’s support of Israel’s ongoing bombing campaign in Gaza following Hamas’ Oct. 7 attack. 

That’s according to a new dispatch from the Institute for Strategic Dialogue (ISD), a non-profit that studies disinformation online. Over the past few years, the CCP has promoted anti-American memes and information online. Spamouflage refers to a disinformation campaign where a state actor uses bots to mass post memes and propaganda across social media. China has ramped up its spamouflage activities in the past few years. Its targets include Canadian Prime Minister Justin Trudeau and the U.S. In particular, the Chinese Communist Party loves posting memes that depict Joe Biden as a deranged and warmongering drug addict who is out of step with the desires of the majority of Americans. 

It’s a pernicious problem. In April, the Justice Department arrested 34 people on charges related to operating one of these troll farms. In January, Google deleted thousands of fake videos on YouTube it said were related to Chinese spamouflage campaigns. In August, Meta killed more than 7,000 accounts it said were related to disinformation campaigns operating against the U.S. Not to be outdone, the U.S. has also been caught operating similar campaigns targeting Russia, Iran, and China. 

According to ISD, the Chinese-linked groups are in a rebuilding phase on X—where most of the activity is happening—as well as on Facebook and YouTube. “After the August takedown on Facebook, Spamouflage appears to have fallen back on accounts which may have flown under Meta’s radar, such as accounts which were active in early 2023 but not active in later months leading up to August,” ISD said in its post. “The campaign also began creating new accounts, for example a series of accounts using AI-generated women’s faces as profile pictures.” 

Washington’s staunch support of Israel’s “siege” on Gaza responding to Hamas’ Oct. 7 attack—Israel has killed nearly 20,000 Palestinians so far, while Hamas killed 1,200 Israelis—have provided China with the raw fuel it needs to create anti-American memes. 

Some of the images appear designed to drive a wedge into the existing divide between Americans who support Israel and Biden’s stance, and those who don’t. In one image, a frightening Uncle Sam sits atop a throne made of guns above the words “American Arms Dealer.” In another, Biden appears in the uniform of an operator in front of an American flag and scenes from Gaza. “Biden Slayer,” the image says. “U.S.’s sole supporter of the war.”

Many of the memes bear the hallmarks of being hastily created with an AI image generator, though it’s not clear which tool was used. Earlier this year, the loose guardrails on both Bing’s image generator and Facebook’s sticker making AI allowed users to create images of corporate mascots participating in atrocities. “These AI images do not appear to be intended to fool anyone into thinking they are real photographs; instead they seem to be intended as graphic illustrations of the key narratives,” ISD said.

According to ISD, these shitposts follow the same pattern as those from previous months. “While the Gaza conflict features prominently, the focus of Spamouflage in recent weeks is clearly shifting towards the US’s 2024 Presidential elections,” it said. “In this context, the operation may start seeking to associate Biden specifically with costly US military entanglements overseas. Depending on how the position of the Republicans and of the likely Republican candidate Donald Trump evolves, this could serve as a wedge issue which the CCP may seek to exploit.

In August, around the time of its massive takedown, Meta published an Adversarial Threat Report detailing these Spamouflage campaigns. “We continue to take down these unsuccessful attempts by Spamouflage, a persistent influence operation, to target audiences in various countries, as part of our ongoing work against coordinated inauthentic behavior,” a Meta spokesperson told Motherboard. “While this campaign struggles to build audiences among authentic communities across over 50 different apps it targets, we’ll keep monitoring and evolving our defenses to stay ahead.”

YouTube has worked with Google over the past few years to identify and remove accounts linked to Chinese-run spam campaigns, which it calls “Dragonbridge.” After the ISD report dropped, it removed 12 accounts related to this activity from the platform. “We have invested heavily in robust systems to proactively investigate coordinated influence operations and terminate their channels and accounts,” a YouTube spokesperson told Motherboard.

X, meanwhile, responded to Motherboard with a boilerplate email saying, “Busy now, please check back later.” X’s moderation policies have been under close scrutiny since billionaire Elon Musk purchased the site and whittled its staff down to a fraction of its original numbers, including moderation teams. Motherboard recently found that the site had not placed restrictions on content promoting apps that create non-consensual AI porn, despite TikTok and  Meta-owned Instagram doing so. 

UPDATE 12/20/23: This story has been updated with a comment from YouTube.