OpenAI published research this week identifying two clusters of ChatGPT accounts, likely run from China, that it says used the chatbot to feed covert influence campaigns into American debates over artificial intelligence. The first, which OpenAI calls "Data Center Bandwagon," generated posts and images claiming that the buildout of AI data centers was driving up household electricity bills. The second, "Tech and Tariffs," produced cartoons attacking US tariffs as a bid for technological dominance, with prompts that specifically instructed the model to leave Xi Jinping out of the picture and focus only on Donald Trump.
The report is careful in a way the headlines around it have not been. OpenAI says it found "no evidence of meaningful breakout" beyond the accounts that created the content, and that what matters is not impact but the fact that PRC-linked operators are testing narratives against AI infrastructure at all. Ben Nimmo, who leads the company's influence investigations, was equally measured: the campaigns were caught precisely because the operatives used ChatGPT to build them, and there is no sign they shifted public opinion at scale.
That nuance is getting lost. OpenAI, several Republican lawmakers and a number of tech investors have folded the findings into a broader claim: that the grassroots opposition to data center construction springing up from rural Virginia to suburban Arizona is, at least in part, a foreign operation. Researchers who actually study those movements told Wired they see no such thing. Local resistance to data centers predates any talk of Chinese interference and is rooted in tangible facts on the ground. A single large facility can draw as much power as a small city and millions of gallons of water a day, and in places like Loudoun County, Virginia, residents have been fighting expansion for years over grid strain, noise and property values.
This is where the story turns from a disinformation report into something more uncomfortable. Foreign actors really do amplify American divisions, and OpenAI's evidence on that narrow point looks solid. But stretching it to explain organic, town-hall-level opposition does something convenient: it recasts citizens worried about their water and power bills as unwitting pawns, and it strengthens the case for federal legislation that would override local zoning in the name of national security. The same move was made over TikTok and over chip export controls. Each time, "China is behind this" turns a messy domestic argument into a test of patriotism.
The backdrop is a Washington that has rarely been more unified on China. At a Senate Banking Committee hearing this week, Chairman Tim Scott and ranking member Elizabeth Warren agreed on little except that the United States cannot let Beijing gain an edge in AI, with Warren warning against "millions more American chips" flowing to China. In that climate, a report about a few hundred inauthentic accounts can do a lot of work. The risk is not that the influence campaign succeeds. By OpenAI's own account, it did not. The risk is that the response treats every inconvenient objection to the AI buildout as something a foreign adversary planted, and stops listening to the people who live next to the substations.