X Found 200,000 Fake Chinese Accounts Trying to Sway Americans Against AI Data Centers

X's Safety team has found a network of 200,000 fake accounts from China on its platform. Some of these accounts were posting content designed to shift how Americans think about AI data centers and energy policy. X shared its findings on its Global Government Affairs page on August 28, 2026 (Engadget).

Out of the 200,000 fake accounts, about 200 were actively posting comic strips, images, and text about AI data centers. X said this content was built to manipulate a real public debate about American AI and energy policy (Engadget).

The posts focused on how AI data centers can raise household electricity bills and put pressure on local power grids. Some accounts shared AI-made cartoons showing data center operators as greedy businessmen getting rich while ordinary families pick up the tab (Engadget).

The fake accounts used OpenAI's tools, including ChatGPT, to create both the illustrations and the talking points. The drawings matched ones described in an OpenAI report from June 2026, which identified China-based accounts using ChatGPT to produce anti-data-center content (Engadget). OpenAI called the activity an attempt to "exploit and amplify existing public concerns" about energy prices connected to AI data centres (Al Jazeera).

OpenAI's June report also found that China-linked ChatGPT users asked the chatbot to write insults aimed at Chinese dissidents and political commentators, as well as criticisms of US tariffs and tech policies. The same accounts posted real news stories about power grid operations and data center demand alongside their fake content. This mix of real and fake made it harder to tell what was genuine public discussion and what was coordinated manipulation (Engadget).

The campaign hit on a real problem. Goldman Sachs reports that US electricity prices rose 6.9% compared to a year earlier as of February 2026, and expects prices to keep climbing as AI facilities demand more power (CNBC). Households near data center clusters genuinely feel the strain, which is exactly why the issue works well for influence operations trying to widen existing divisions.

This is not a one-off. In July 2024, the United States and its allies shut down a Russian bot farm that used AI to spread propaganda to Americans (CSIS). A research paper from the Bitcoin Policy Institute documents a broader multi-year foreign influence campaign aimed at slowing US AI development (Bitcoin Policy Institute).

What makes this case notable is the layered approach. AI-generated cartoons and talking points zeroed in on a live, contested policy debate, then mixed in real news to give the campaign credibility. The people behind it were not inventing the problem. Electricity prices are going up. Power grids are under strain. The manipulation came from amplifying one side of the story, adding emotionally charged images, and pushing people toward a specific conclusion about who is to blame.

There is also a tricky challenge for AI companies here. OpenAI's own tools were used to create the content that OpenAI's later report flagged as fake. The same technology that makes it easier for anyone to create content also makes it cheaper to fake public opinion. Detection is getting better, but it still lags behind. OpenAI's report surfaced the campaign in June, and X expanded the account-level picture two months later. By then, the influence content had already been circulating.

For anyone following this space, the practical takeaway is straightforward. Foreign influence operations targeting US AI infrastructure policy are now a recurring pattern. They use real grievances, commercial AI tools, and genuine policy debates where the facts are contested. Telling real public concern apart from coordinated manipulation will take platform-level detection, AI company threat reporting, and reader-level media literacy all working together. No single layer is enough.