The irony is almost too perfect: China-linked actors used American AI to argue against American AI infrastructure.
The Summary
- OpenAI banned ChatGPT accounts linked to China after detecting them running an influence operation to generate opposition to US data centers
- The campaign aimed to exploit existing concerns about energy prices and local impact to hinder US competitiveness in AI infrastructure
- The operation gained little authentic engagement on X and Facebook, staying small-scale and short-lived
- This is the first time OpenAI has caught Chinese influence operations specifically targeting US data center debates using their own models
The Signal
OpenAI's Intelligence and Investigations team identified what they're calling the "'Data Center Bandwagon' Campaign", a suspected Chinese operation that used ChatGPT to generate content aimed at fueling local opposition to data center construction across the US. The operation created fake accounts posing as Americans and jumped into pre-existing local debates about data centers. Ben Nimmo, the principal investigator on the case, noted the particular irony: "Under the circumstances, it's particularly ironic that they use American AI to do it."
This marks a tactical shift in how nation-state actors approach AI infrastructure competition. Rather than directly attacking AI labs or stealing models, the play here was to slow down the physical infrastructure buildout that makes frontier AI possible. Data centers are the chokepoint. You can have the best models in the world, but without compute capacity at scale, you're stuck.
"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."
The operation strategy was smart, even if the execution was clumsy:
- Target real community concerns about energy prices and local environmental impact
- Amplify existing skepticism about big tech's resource consumption
- Frame opposition as grassroots American pushback rather than foreign interference
Bloomberg reports that OpenAI sees this as an attempt to hinder US competitiveness in AI, echoing broader industry concerns about infrastructure as the new battleground. The good news is the operation flopped. Most posts gained little to no authentic engagement, suggesting the personas weren't sophisticated enough or the targeting was off.
What's more significant is the detection method. OpenAI caught this because the operators relied on ChatGPT to generate their content. That means OpenAI could see usage patterns, content similarities, and account clustering that revealed the operation. This is the first public case of China-linked actors using OpenAI's models to target US data center debates specifically, though OpenAI has identified other Chinese influence operations before.
The Implication
Expect more of this. Data center opposition is a real political issue in communities across the US, which makes it perfect cover for influence operations. The next attempts will be harder to detect because operators will learn not to use the same AI tools for content generation that their targets control. Watch for similar campaigns using open-source models or other providers where detection is harder.
For AI companies and policymakers, this is a preview of infrastructure competition in the agent era. The race isn't just about who builds the best models. It's about who can build the compute capacity to run them at scale, and whether local opposition, whether organic or manufactured, can slow down one side more than the other.