OpenAI Says China-Linked Accounts Are Seeding Opposition to US Data Centers
A new OpenAI report details PRC-linked influence operations using ChatGPT accounts to stir up local resistance to AI infrastructure in the United States.
Crédit photo: Image via Bloomberg — Technology. Used under fair use for news commentary. · source
OpenAI has identified a network of ChatGPT accounts with ties to China that it says have been working to amplify local opposition to data centers across the United States, potentially as a way to slow American AI development.
That's a striking claim. And honestly, it's one that deserves more scrutiny than it's getting.
The report, published on the OpenAI Blog, describes a coordinated influence operation it's linking to the PRC. The accounts were apparently using AI-generated content to target U.S. tech debates, narratives around data centers, tariff policy, and, weirdly, spreading false claims about ChatGPT itself. That last part is almost funny, in a bleak way. Using ChatGPT to spread lies about ChatGPT.
According to Bloomberg, the operation was specifically trying to stir up grassroots-style pushback against data center construction in American communities. The framing OpenAI uses is that this could hinder U.S. competitiveness in AI, by making it harder to build the physical infrastructure the whole industry runs on.
I initially thought this sounded like the kind of vague, hard-to-verify geopolitical claim that tech companies sometimes make when it's convenient. But the specificity here, targeting data center opposition at the local level, is actually pretty interesting.
À lire aussi
More in AI Models
A former xAI engineer says he was let go for questioning Grok's safety guardrails. If the allegations hold up, it's a familiar and depressing pattern.
Robert "Bob" Macintosh · 4 hours ago · 4 min
With Anthropic and OpenAI eyeing IPOs and billions flowing into AI infrastructure, a UCLA professor is asking the question the industry keeps sidestepping.
Robert "Bob" Macintosh · 8 hours ago · 4 min
Oracle's capital expenditure forecast is eye-watering, and the market's reaction suggests people aren't sure the AI infrastructure bet will pay off.
Sarah Williams · 10 hours ago · 3 min
Middle Eastern sovereign wealth funds are piling into the SpaceX IPO, and Goldman Sachs thinks the whole thing is a signal about where serious money wants to go next.
You might be wondering why anyone would bother. Data centers aren't exactly a hot-button culture war issue. But that's sort of the point.
Data centers are genuinely controversial in a lot of American communities right now. They consume enormous amounts of water and electricity. They create relatively few local jobs for their footprint. Residents in places like Virginia, Iowa, and Nevada have raised real, legitimate concerns about them for years.
So if you wanted to slow down U.S. AI infrastructure without doing anything obviously aggressive, amplifying existing local opposition is actually a pretty clever approach. You're not inventing the concern. You're just turning up the volume on something that's already there.
That's what makes this difficult to assess. The underlying community concerns are real. The question of whether they're being artificially amplified is a separate one, and it remains unclear exactly how much reach or impact these accounts actually had. OpenAI's report doesn't appear to quantify that in a way that's easy to evaluate.
Bloomberg notes that OpenAI's framing here echoes recent rhetoric from others in the tech industry. That's worth sitting with for a second.
There's a version of this story where a genuine foreign influence operation is being exposed. There's also a version where a company with enormous financial interest in data center expansion is framing local opposition as a national security threat. Both things can be partially true at the same time, tbh.
I'm not saying OpenAI is wrong. I'm saying the incentive structure here is complicated, and I think readers should hold that in mind. When a company that needs to build a lot of data centers publishes a report saying opposition to data centers is being fueled by foreign adversaries, that's not automatically false, but it's also not automatically neutral.
The U.S. is in a real race to build AI infrastructure. The data center bottleneck is a serious constraint on domestic AI development, and there's bipartisan political will to address it. At the same time, communities have legitimate environmental and resource concerns that deserve to be heard and taken seriously, not dismissed as foreign manipulation.
If influence operations are genuinely making it harder to have that conversation honestly, that's a real problem. But the solution can't just be "trust the tech companies and build faster." That's too convenient.
What we don't know yet is how sophisticated or effective this particular operation was. OpenAI identified the accounts. It doesn't tell us, at least not in detail I could find, how many people actually saw the content, whether it changed any minds, or whether it had any measurable effect on local policy decisions. This is based on what OpenAI has chosen to disclose, which is a limitation worth naming.
This isn't the first time OpenAI has published reports on AI-assisted influence operations. The company has been fairly consistent about flagging these, which I think is genuinely useful. Foreign actors using AI tools to run influence campaigns is a real and growing concern, and transparency about how those tools are being abused matters.
But each report also lands in a specific political moment, and this one lands in a moment when the U.S. tech industry is pushing hard for fewer regulatory barriers to data center construction. That context shapes how the report gets used, regardless of what OpenAI intends.
Honestly, I think the most important thing here isn't the specific influence operation, which may or may not have had much real-world effect. It's the underlying dynamic it reveals. The physical infrastructure of AI, the land, the power, the water, the permits, is becoming a geopolitical battleground in ways that are genuinely new. That's the story I keep coming back to.
We're used to thinking about AI competition in terms of models and chips and talent. The fight over who gets to build data centers, and where, and how fast, is going to matter just as much. And it's a fight that's now apparently happening at the level of local planning meetings and community Facebook groups.
