OpenAI Finds China-Linked ChatGPT Push on US AI Debates

OpenAI says it banned two clusters of ChatGPT accounts it believes were run by a private Chinese technology company working for provincial-level government clients, accusing the operators of running covert influence operations aimed at US debates over AI data centres and tariffs. The company laid out the findings in its June 2026 threat report, published Wednesday, and concluded the social media effort received “no authentic engagement” and showed “no evidence of meaningful breakout beyond its own activity.”

The Two Operations OpenAI Banned

OpenAI said it had identified and removed two clusters of ChatGPT accounts likely originating from China, both tied to attempts to push content into US political conversations. The first, which the company dubbed the “Data Centre Bandwagon” campaign, generated social media comments and images claiming that AI data centres were driving up electricity prices for ordinary Americans. The second, the “Tech and Tariffs” campaign, produced comments and political cartoons criticising US tariffs and portraying Washington as seeking technological dominance.

Ben Nimmo, OpenAI’s principal investigator on the intelligence and investigations team, framed the cases as classic attempts to ride a domestic argument that already existed. “This looks like a classic example of a foreign influence operation, jumping onto the bandwagon of a genuine pre-existing domestic debate, and trying to manipulate it by using fake accounts, posing as Americans,” Nimmo told reporters on a Wednesday call ahead of the report’s release. He added: “Under the circumstances, it’s particularly ironic that they use American AI to do it.”

OpenAI’s report, titled “PRC-linked influence operations are targeting AI debates in the US,” is the first time the company has tied a China-linked operation to its AI models and the data centre debate. The document details how both clusters used likely inauthentic social media accounts, sought English-language output despite prompting in Simplified Chinese, and relied on virtual private networks to reach ChatGPT, which is not officially accessible from mainland China. The full report is available on OpenAI’s threat intelligence site.

Inside the “Data Centre Bandwagon” Campaign

The Data Centre Bandwagon campaign was the larger of the two by OpenAI’s reckoning, and the only one for which the company offered a specific institutional attribution. OpenAI wrote in the report that the users were “likely part of a social media operations team at a private Chinese technology company conducting work for Chinese provincial-level government clients.” The accounts prompted ChatGPT in Simplified Chinese while repeatedly asking for English- and Chinese-language outputs, then posted the results to X and Facebook via likely inauthentic profiles.

Visual content was central to the operation. OpenAI said the campaign produced comic strips and image posts about power grid capacity and electricity prices, often paired with links to legitimate news coverage of data centre power demand. The goal, in the company’s description, was to amplify existing American concerns about energy costs and the local impact of new AI infrastructure, not to invent them.

The same networks also spread false claims that ChatGPT user data had been compromised. OpenAI called those allegations false, in quotes the report printed directly. “These allegations were entirely false,” the company said. The line sat in the report alongside the data centre material, suggesting the operators were running more than one narrative in parallel.

Nimmo told reporters the campaign’s run was short and small. “This was not a case of an influence operation creating a debate. The debate existed already. This was an influence operation from China trying to interfere in it. We didn’t see any signs that they succeeded,” he said on the call. Most of the posts gained little to no traction, and OpenAI found no evidence of meaningful breakout beyond the accounts’ own activity.

The “Tech and Tariffs” Operation and Its Posters

The second campaign was narrower in scope and harder to pin down. OpenAI said a separate cluster used ChatGPT to generate short comments and political cartoons that criticised US tariffs and the US push for global tech dominance. The company said it could not link this campaign to a particular entity in China, and it stressed that the campaign’s reach was equally thin.

The visual artefacts the campaign produced were pointed and specific. OpenAI included several examples in the report. They included:

• A cartoon of Trump sawing through a ladder while standing on it, framed as a critique of his own tariff policy.
• A cartoon of Trump wearing American flag-patterned pants reading “America First” and swinging a mallet labelled “Tech Dominance” into a wall reading “Global Future.”
• Comments and posts that pressed the line that Washington was pursuing tech dominance at the expense of ordinary consumers.

One detail stood out to OpenAI’s investigators. The campaign’s operators specifically instructed the system to depict only President Donald Trump and to avoid references to Chinese President Xi Jinping, a constraint Nimmo’s team read as a deliberate choice to keep the propaganda deniable. The posts clustered around October 2025, when Trump announced an additional 100% tariff on Chinese goods, an alignment OpenAI flagged in the report as notable.

The Debate the Campaigns Tried to Hijack

What the OpenAI report lands in is a domestic argument that was already running hot. A Gallup poll cited in coverage of the report found that 71% of Americans somewhat or strongly oppose construction of data centres in their communities. A separate Harvard/MIT poll found 32% of Americans oppose data centres in their area while 40% support them. On tariffs, a Harris poll released in March found that seven in 10 Americans said Trump’s tariffs had caused them to pay higher prices.

Industry-side data underline the same trajectory. According to Data Center Watch, local opposition blocked or delayed dozens of US data centre projects in 2025, representing more than $150 billion in potential investment. Senator Bernie Sanders has called for moratoriums on new facilities, and Inside the data centre fight tracks the politics behind the building boom. The political temperature around new AI buildouts is high on its own, and OpenAI’s report landed in the middle of it. For related reading on how US politicians are pressing AI executives on the China question, see Nvidia CEO Jensen Huang’s Senate testimony on China chip export controls.

Who Else Is Saying the China Line

Rep. John Moolenaar, the Republican chairman of the House Select Committee on the Chinese Communist Party, picked up OpenAI’s report within hours and pointed it at a broader argument. “There are legitimate questions about data centres, and Americans deserve answers as companies work to build the infrastructure we need for our nation’s future,” Moolenaar said. “Unfortunately, the Chinese Communist Party exploits our openness and works to divide Americans through its United Front organisations and other entities.” He added that some elements in China were attempting to influence the data centre debate “by more disguised methods,” and said his committee would keep investigating.

Other Republicans have gone further in public. Congressman Brett Guthrie of Kentucky, who chairs the House Energy and Commerce Committee, sent a letter last week to FBI Director Kash Patel and to White House advisers David Sacks and Michael Kratsios, calling for a briefing on whether the US government has evidence of Chinese influence in the data centre debate. Several Republican members of Congress also sent a separate letter to the Trump administration this month raising alarms about “foreign influence campaigns working to slow American AI progress.”

Outside Congress, the China-labelling has come from tech investors with stakes in the buildout. On the All-In podcast, fund manager Gavin Baker said it was starting to feel “like there might be a CCP-funded campaign” against data centres. Kevin O’Leary, the Shark Tank star and investor in the Stratos data centre project in Utah, claimed on Fox News that he had found two cells inside of Utah working against the project and named two local groups, Alliance for a Better Utah and Elevate Strategies, as Chinese-government fronts.

Both groups denied the claim. Gabi Finlayson, a senior partner at Elevate Strategies, told NPR she had no idea how she or her firm ended up on O’Leary’s list. “We are certainly not a Chinese cell. Nobody pays us to make any content, let alone any foreign government,” she said. Alliance for a Better Utah’s communications director, Elizabeth Hutchings, said the organisation posted roughly $200,000 in revenue in 2024, in line with the prior decade, and called O’Leary’s allegations an “outrageous claim” that left her staff laughing.

Independent researchers tracking foreign influence online are more cautious. Darren Linvill, co-lead of the Media Forensics Hub at Clemson University, told NPR his team had turned up little. “We haven’t found much,” Linvill said, adding that Chinese state media appeared far more interested in promoting Chinese data centres than in opposing American ones. A mid-May report from the Bitcoin Policy Institute claimed “three vectors of influence” behind anti-AI activism, but a Washington Post follow-up found several of the named groups denied Chinese funding and the report itself offered little direct evidence of a coordinated campaign.

Why the Report Travels the Way It Does

OpenAI’s own framing of the report cuts both ways. The company warned that the themes it identified are “likely to remain attractive” for Chinese influence operations because they can be “inserted into legitimate public debates while nudging audiences toward distrust of US institutions.” That is also exactly the framing that US politicians, tech investors, and pro-buildout commentators are using in real time to characterise a much larger, mostly domestic backlash to new data centres. The report’s first-of-its-kind finding, that AI models are now part of the toolkit for these campaigns, gets deployed alongside claims about the broader debate without the rest of the report’s caveats.

OpenAI is not a neutral messenger on the underlying issue. The company is pushing aggressively for new data centre construction to meet surging demand for its own products, and has publicly cast AI infrastructure as a matter of national competitiveness with China. A report that documents a small, unsuccessful foreign influence operation, even one that is honestly reported, also serves an agenda that wants the domestic opposition questioned. As Nimmo himself said, the real grievance on energy costs and local impact is real, foreign amplification or not.

For a wider view of how Chinese AI players are scaling against the US, the picture is not a one-way street. Moonshot AI’s $30 billion funding sprint shows Chinese model companies raising capital at a pace that US hawks are watching closely. The politics around US data centres will keep running through both stories, the foreign influence one and the industrial competition one, for the rest of the year.

Frequently Asked Questions

What did OpenAI’s June 2026 threat report find?

OpenAI said it banned two clusters of ChatGPT accounts it described as likely originating from China. The first, the “Data Centre Bandwagon” campaign, generated English-language posts and comic strips arguing that AI data centres were driving up electricity prices in the United States. The second, the “Tech and Tariffs” campaign, produced comments and political cartoons criticising US tariffs and the US push for tech dominance. OpenAI concluded that neither campaign achieved meaningful reach on social media.

How did the campaigns try to influence US debates?

Operators used ChatGPT to draft content, then posted it through likely inauthentic accounts on X and Facebook. The Data Centre Bandwagon campaign paired generated commentary with links to legitimate news coverage of data centre power demand. The Tech and Tariffs campaign produced political cartoons, including images of President Trump sawing a ladder he was standing on, while instructing the model to avoid depicting Chinese President Xi Jinping.

Did the campaigns actually work?

OpenAI said no. The company said the posts received “no authentic engagement” and that it found “no evidence of meaningful breakout beyond its own activity.” Ben Nimmo, who leads threat investigations at OpenAI, told reporters, “We didn’t see any signs that they succeeded.” The Data Centre Bandwagon campaign was described in the report as small scale and short in duration.

Why are some US politicians and commentators pushing this narrative?

Rep. John Moolenaar, chairman of the House Select Committee on the Chinese Communist Party, framed the findings as evidence of a broader Chinese effort to divide Americans on data centre policy. Tech investors with stakes in data centre projects, including Kevin O’Leary and Gavin Baker, have also publicly suggested that local opposition to new facilities is being driven or funded from China. Several Congressional Republicans have asked the Trump administration for a briefing on the question.

Is there evidence of broader Chinese influence in the US data centre opposition?

Independent researchers say not much so far. Darren Linvill, co-lead of the Media Forensics Hub at Clemson University, told NPR, “We haven’t found much.” A mid-May report from the Bitcoin Policy Institute claimed “three vectors of influence” behind anti-AI activism, but offered little direct evidence of a coordinated Chinese effort, and the groups it named have denied any Chinese funding. A Gallup poll cited in coverage of the debate found that 71% of Americans oppose data centre construction in their communities, a baseline of organic opposition that predates the OpenAI report.