OpenAI's June 10 threat report shows China-linked operators using ChatGPT to test anti-data center narratives in the United States, turning a local infrastructure fight into a geopolitical risk investors can no longer ignore.
The irony is hard to miss: an AI company says China-linked operators used an AI chatbot to attack the data centers that make modern AI possible. OpenAI's June 10 threat report described two influence campaigns that used ChatGPT to generate posts, cartoons, and comments aimed at American debates over technology policy, tariffs, electricity prices, and data center construction. The campaigns did not break through. That is the reassuring part. The less reassuring part is that they targeted a real and growing weakness in the US AI buildout.
The first operation, which OpenAI called "Data Center Bandwagon," focused on the pressure points already worrying American communities: power bills, grid strain, land use, and the local impact of giant server campuses. OpenAI said the users behind it appeared likely to be part of a social media operations team at a private Chinese technology company working for provincial-level government clients. They prompted ChatGPT in Simplified Chinese and asked for English and Chinese outputs, then used the model to create material criticizing US data centers.
That detail matters because this was not a generic anti-tech campaign. It was built around a specific political vulnerability. Data centers have moved from back-office infrastructure to neighborhood controversy, especially in places where residents are already asking whether AI demand will raise utility costs or overwhelm local planning. Operators did not have to invent that argument. They only had to borrow it, polish it with AI, and push it through accounts meant to look like ordinary people.
The second campaign, "Tech and Tariffs," used ChatGPT for short comments and political cartoons criticizing US technology policy and tariffs. According to Business Insider's account of OpenAI's findings, the campaign was active around Donald Trump's October 2025 announcement of an additional 100% tariff on Chinese goods, and the users specifically asked that Xi Jinping be kept out of cartoon images. That instruction tells you a lot about the intended audience. The content was not designed to criticize China. It was designed to land on American screens and frame US technology strategy as reckless or aggressive.
OpenAI's assessment was clear: neither operation gained meaningful traction. The data center campaign was small, short lived, and produced little to no authentic engagement on platforms including X and Facebook. The company said it found no evidence of meaningful breakout beyond the operators' own activity, and it terminated the relevant accounts. By the usual measure of influence operations, that makes these campaigns failures.
But the target is more important than the performance. Data center opposition in the United States was already becoming a serious constraint before this report arrived. The Washington Post recently reported, citing Data Center Watch, that at least 48 data center projects were blocked or delayed across the country in 2025, affecting $156 billion in planned development. The same coverage said grassroots organizations opposing data centers have grown to nearly 400 nationwide. That is not fringe resistance. It is a permitting and political problem with real capital consequences.
This is where the story moves from disinformation cleanup to infrastructure risk. The facilities being criticized are the same facilities needed to train and run large language models, including the kind of model used to generate the campaign material. Every AI-generated post about electricity bills depends on compute capacity, power contracts, transmission access, and local permission. A foreign-linked operator using ChatGPT to weaken support for AI infrastructure is not a future scenario. OpenAI says it has already seen the playbook.
There is also a clear geopolitical logic. China does not need to stop US AI infrastructure outright to benefit. Delays are enough. A rejected zoning application, a postponed interconnection agreement, or a bond market that starts pricing in higher political risk can slow the buildout that American AI companies need. Foreign influence campaigns do not need to manufacture public anger from nothing. They are more useful when they amplify arguments that already exist.
Investors should pay attention because the numbers around AI infrastructure are now enormous. Moody's Ratings recently forecast that Amazon and other large AI hyperscalers could spend $785 billion this year and nearly $1 trillion in 2027, much of it tied to data center expansion. That scale changes the risk model. Community opposition used to look like a local planning headache. The OpenAI report gives it another layer: the possibility that foreign actors will try to intensify local resistance around projects that are strategically important to US technology leadership.
The practical lesson is not that every data center critic is part of a foreign campaign. That would be lazy and wrong. Many residents have legitimate concerns about electricity costs, water use, noise, land development, and tax incentives. The sharper point is that real grievances are exactly what influence operators look for. They work best when they can blend into an existing debate and make it louder, angrier, or more confusing.
OpenAI's report gives developers, policymakers, and financiers something concrete to watch. The next campaign may be better written, better targeted, and harder to separate from authentic local frustration. Data center companies will still need to solve the ordinary problems: grid capacity, permitting, community benefits, and transparency. But they now have to treat the information environment as part of the project site. The fight over AI infrastructure is no longer only happening in planning meetings. It is also happening in feeds, comments, cartoons, and posts designed to make a real bottleneck worse.
Also read: Fresh US strikes on Iran are squeezing metals on two tracks that traditional safe-haven logic cannot explain • Opendoor shut its India office and Wall Street handed it an 8% stock gain that every startup board will remember • TDK's $400 million acquisition of Fabric8Labs proves that keeping AI chips cool has become as strategically important as building them
