Two Coordinated Clusters Target U.S. Tech and Energy Policy Debates

OpenAI has disclosed that it banned two distinct clusters of ChatGPT accounts linked to operators in China, which were conducting covert influence operations aimed at manipulating American public discourse around technology policy and energy costs. The company's latest threat intelligence report reveals a sophisticated playbook: operators used ChatGPT to generate AI-crafted political cartoons and social media commentary, then deployed them on X (formerly Twitter) under fabricated American personas.

The first cluster, dubbed "Data Center Bandwagon," produced comic strips and social media posts blaming AI data centers for rising household electricity bills. The second, "Tech and Tariffs," generated anti-tariff cartoons depicting President Trump while conspicuously avoiding any portrayal of Chinese leader Xi Jinping. Both clusters prompted ChatGPT in Simplified Chinese via VPNs and posed as Americans from diverse backgrounds on X.

Data Center Bandwagon: Weaponizing Real Energy Concerns

The Data Center Bandwagon operation centered on a genuine and well-documented controversy: PJM Interconnection's independent market monitor has attributed an "irreversible" 75.5% increase in capacity auction prices across the largest U.S. grid region to data center demand growth. Wholesale prices near some data center clusters have climbed as much as 267% in five years, prompting three U.S. senators to demand answers from Amazon, Google, and Meta over costs passed to residential customers.

Operators exploited this legitimate debate. They asked ChatGPT to create comic strips about grid operator capacity auction prices, drawing on regional newspaper reporting, then posted the AI-generated output on X under hashtags like #capacityauction alongside links to legitimate news coverage. OpenAI assessed the operators were likely a social media team at a private Chinese tech company working for provincial-level government clients.

"This was not a case of an influence operation creating a debate," said Ben Nimmo, principal investigator at OpenAI. "It was an influence operation attempting to amplify and distort an existing debate."

OpenAI rated the activity Category One on the Breakout Scale, meaning it remained confined to a single platform with no evidence of reaching authentic audiences.

Tech and Tariffs: A Multi-Language "Water Army"

The second cluster, Tech and Tariffs, took a broader geopolitical aim. It generated anti-tariff cartoons under explicit instructions to depict President Trump but never Xi Jinping, and was linked to a network of fake X accounts spreading fabricated claims that ChatGPT user data had been compromised—a tactic OpenAI believes was intended to damage its own reputation.

This cluster produced bulk comment batches in English, Italian, Japanese, and Traditional Chinese (the latter targeting audiences in Taiwan). One operator explicitly described the accounts as a "water army"—a Chinese term for coordinated troll networks—and even asked ChatGPT to design a system for scraping and analyzing social media posts from individuals flagged as risks. OpenAI's model declined the request, returning only generic data storage advice.

Historical Parallels: The Spamouflage Playbook

OpenAI's threat report draws a direct line to the 2022 Spamouflage operation, which researchers at the Australian Strategic Policy Institute (ASPI) and Mandiant found targeting Lynas Rare Earths, Appia, and USA Rare Earth after Beijing's 14th Five-Year Plan prioritized rare earths. The new activity followed the adoption of the 15th Five-Year Plan recommendations, which elevated AI as a strategic industry for China.

The parallel is instructive: both operations used state-aligned actors to shape narratives around strategic resources—rare earths then, AI infrastructure and energy policy now—through coordinated inauthentic behavior on Western social platforms.

Implications for AI Safety and Platform Governance

The case underscores a growing challenge for AI providers: the same generative capabilities that enable creative expression can be weaponized for narrative engineering at scale. OpenAI's decision to publish detailed threat intelligence—including the specific prompts used, the operational infrastructure (VPNs, Simplified Chinese prompts), and the platform-level attribution—sets a transparency benchmark for the industry.

Yet the Category One rating also highlights a critical gap: influence operations that fail to "break out" to authentic audiences may still pollute information environments, seed narratives for later amplification, and erode trust in legitimate policy debates. The Data Center Bandwagon campaign didn't create the energy cost debate—it hijacked it.

As AI-generated content becomes indistinguishable from human output, the burden shifts to platforms, policymakers, and the public to develop new verification frameworks. OpenAI's disclosure is a necessary step, but not a sufficient one.