China often tests its influence campaigns on Taiwan before turning them on other democracies. With both Taiwan and the U.S. voting this month, researchers say the same AI-powered playbook is at work in each: seize on real grievances and deepen distrust rather than pick winners.
Taiwan’s National Security Bureau told lawmakers last month it had tracked more than 2.5 million pieces of online misinformation from China from January to September, up from about 1.69 million a year earlier.
China is “expanding the use of AI technology for information manipulation” to sway voters and election results, the intelligence agency said. Beijing has repeatedly denied interfering in Taiwan’s elections.
The threat is already in the U.S.
In June, an Associated Press investigation found that between 2023 and 2025, DEA agents had let fentanyl shipments reach New Mexico streets. New Mexico politicians took to social media to criticize the Trump administration. The Chinese influence operators saw an opportunity.
Since Sept. 22, accounts linked to Spamouflage, an influence operation linked to China, have piggybacked on the politicians’ posts on X, according to Graphika, a New York-based firm that tracks online influence operations.
The accounts replied to posts by Gov. Michelle Lujan Grisham, Sen. Ben Ray Luján and Reps. Melanie Stansbury, Teresa Leger Fernández and Gabe Vasquez, blaming the Drug Enforcement Administration for the state’s fentanyl crisis. All five are Democrats, and the four members of Congress are running for re-election.
Graphika said that the activity was likely an effort to undermine confidence in U.S. institutions before the midterms.
The campaign also deflects attention from China’s own role in the opioid crisis. Washington has long accused Beijing of failing to stop the flow of chemicals used to make fentanyl. AI-generated images are clearly aligned with China’s state media narratives over the issue.
Few readers appeared to interact with the posts. But Graphika analyst Lili Turner noted that before the 2024 election, one of Spamouflage’s TikTok videos mocking then-President Joe Biden received about 1.5 million views.
The campaigns are “just too important to ignore,” she told MBN.
Real grievances, powered by AI
The New Mexico case fits a broader pattern: Influence operations don’t invent issues; they twist real ones.
The campaigns amplify “legitimate grievances or legitimate societal discord directed at the incumbent party,” rather than Democrats specifically, Turner said.
Ethan Tu, founder of Taiwan AI Labs, a Taipei nonprofit that tracks coordinated online accounts, sees the same tactic in Taiwan.
The aim is to deepen divisions among voters and erode trust in allies, institutions and the media, often by seizing on the cost of living and energy prices, he said.
“In every major election, there are operations,” Tu said.
AI is making them much harder to spot. Spamouflage once posted in clumsy English, but generative AI has smoothed the syntax, Graphika’s Turner said.
Fake accounts used to copy and paste the same text and never reply, Tu said.
“Now, they act like a chatbot,” he said.
Some users have unmasked such accounts by asking them to write computer code, which they did.
Audrye Wong, a senior fellow at the American Enterprise Institute who studies Chinese influence in American state and local political campaigns, said Beijing is “quite deliberately trying to incorporate AI” into its influence operations.
Beyond producing content, she said, China is trying to build systems that gather data on how different audiences think and simulate how they would react to different kinds of Chinese messaging.
She pointed to OpenAI’s June 10 report, which found that accounts likely originating in China used ChatGPT to generate posts and cartoons claiming AI data centers were driving up Americans’ electricity bills. Like the New Mexico campaign, it seized on a real concern but appeared to gain little traction.
But OpenAI said the case was significant because it showed “PRC-origin influence operators testing narratives against AI infrastructure.”
How Taiwan fights back
Taiwan has shifted from debunking false claims to warning the public before they spread. Fact-checking alone stopped working around 2020, Tu said, as operations moved from outright lies to conspiracy theories and panic, claims that are hard to prove true or false. He pointed to Taiwan’s 2023 egg shortage. The shortfall was a real but small dip of a few percent in the daily supply, according to the Council of Agriculture. What operators exploited, Tu said, was the panic behind a different question: “Will I be able to buy eggs tomorrow?”
“For every rumor you debunk, 10 more come out,” Tu said. His team tries to flag misinformation accounts early and alert journalists and policymakers.
Taiwan’s government also reports publicly on the threat. Prosecutors brought 71 cases under the country’s Anti-Infiltration Act during the 2024 elections.
Tu sees Taiwan as an early warning system. China, which views Taiwan as part of its territory, often tests misinformation campaigns there first.
“Taiwan is like a petri dish for democracies,” he said, describing how narratives tested there later appear elsewhere. “We are on the front line of information warfare.”
Tu’s advice was direct: American platforms and political leaders “should come to Taiwan more often to discuss the information warfare Taiwan has faced.”
Why it matters to the Middle East
In the Middle East, China’s messaging follows a different tactic. Graphika has seen Chinese and China-aligned actors portray the Iran war as a loss for the U.S., a source of tension with allies and a threat to global peace.
The campaigns are “more interested in undermining the U.S. on the global stage and signaling support for Iran,” Turner said.
The same message runs through official channels. When a U.S.-Iran ceasefire faltered in June, Chinese state media shifted from neutral coverage to blaming the United States and Israel for renewed fighting, while pro-China commentators cast the deal as a humiliation for Washington, according to Graphika.
Chinese outlets have also amplified Iranian officials, including some from the Islamic Revolutionary Guard Corps, who vow to counter what it called “the enemy’s economic warfare.”
The U.S. State Department has designated the IRGC a terrorist organization.
Chinese and pro-Iran accounts also amplify each other openly, and this messaging travels much further than covert campaigns.
In August, a pro-Iranian regime account on X posted a clip of China’s Foreign Ministry spokesperson rejecting new U.S. sanctions on Iran; it drew about 500,000 views, Graphika found.
Sometimes the open and the covert work together.
Tu said his team has tracked operations in which covert networks first create fake “impartial” news outlets and grassroots supporters, then use the outlets to report on those supporters. Only then do the state media step in.
“After shaping fake public opinion, the state media of China, Iran and Russia began reporting on it,” Tu said.
Taiwan’s National Security Bureau said in its report that China is using regional conflicts to deepen cooperation with Russia, North Korea and Iran.
On their own, these campaigns rarely change minds. But by seizing on real grievances, from the fentanyl crisis to the Iran war, they aim to chip away at trust in the United States, at home and abroad.
