X Uncovers 200,000 Bot Farm Spreading Anti-AI Data Center Propaganda

X has discovered a sprawling bot farm network consisting of 200,000 accounts operating on its platform. The company’s Safety team, as part of an investigation into suspected inauthentic Chinese accounts, announced the findings on its Global Government Affairs page. While the full scope of the operation remains unclear, at least 200 accounts were actively posting comic strips, images, and text about AI data centers in ways designed to manipulate legitimate policy debates.

The Campaign Strategy

The inauthentic accounts focused their messaging on a real concern: the strain that hyperscale AI facilities place on local power grids and electricity costs. Their posts claimed that AI data centers drive up household electricity prices and burden nearby communities. The operation went further by distributing AI-generated cartoons depicting data center operators as greedy businessmen profiting while average families pay the price.

Based on the sample posts X shared publicly, the network relied heavily on OpenAI’s ChatGPT to generate both illustrations and talking points. Interestingly, this mirrors findings from an OpenAI report published in June about similar China-linked accounts that used ChatGPT to create anti-data center content. OpenAI noted at the time that these accounts mixed generated material with actual news stories about power grid capacity and data center power demands, making the disinformation more credible.

The Broader Context

The irony here is that some of these accounts’ core arguments aren’t entirely fabricated. There are legitimate concerns about rising electricity costs in communities near major tech facilities. According to Goldman Sachs, electricity prices jumped 6.9 percent year-over-year as of February 2026, with further increases expected as AI infrastructure demand continues climbing. This makes the disinformation campaign particularly insidious: it exploits real anxieties to push a specific geopolitical narrative.

X stated in response: “We remain committed to maintaining an open and authentic platform where people debate topics of public interest. We take seriously any attempts to undermine the integrity of the global town square and suspend accounts that violate our Authenticity policy.” The company suspended the accounts involved but notably did not release the full list of identities.

What Else Was This Network Doing?

Here’s where things get murkier. X hasn’t revealed what the remaining 199,800 accounts in the troll farm were posting about. The OpenAI June report provides some clues: the China-linked ChatGPT users also requested the chatbot generate insults targeting Chinese dissidents and political commentators. They additionally asked for content criticizing US tariffs and technology policies.

This suggests the bot network was executing a sophisticated tech influence operation far broader than just energy policy arguments. The anti-data center campaign appears to have been one component of a multifaceted disinformation effort.

The AI Generation Problem

The reliance on ChatGPT to generate campaign content reveals an uncomfortable truth: AI tools have become the infrastructure of modern influence operations. Text-to-image models can produce convincing propaganda at scale. Language models can craft compelling narratives in any style or tone. This democratizes disinformation creation, removing technical barriers that once limited such campaigns.

What’s particularly troubling is how difficult it becomes to distinguish AI-generated propaganda from organic discourse when mixed with legitimate reporting and real policy concerns. The OpenAI accounts cited actual news stories alongside generated content, making the fabricated elements harder to spot.

This incident underscores why tech platforms need increasingly sophisticated detection systems. Yet it also highlights a fundamental challenge: can authenticity policies keep pace with AI-powered disinformation? When bot farms can generate thousands of unique posts daily using language and image models, human moderators and even algorithmic detection face an uphill battle.

The real question isn’t whether X will eventually catch these operations, but whether catching them after they’ve already influenced the conversation is enough.

Source: X Global Government Affairs announcement and OpenAI June 2024 report