OpenAI blocks two covert Chinese-linked networks using ChatGPT to inflame organic American debates over tariffs and data centers.
The front lines of geopolitical warfare have firmly shifted into the digital sandbox of generative artificial intelligence. For years, cybersecurity experts warned that bad actors would eventually weaponize large language models (LLMs) to automate and scale up disinformation campaigns. That reality is actively playing out, but with a highly meta twist: foreign operatives are now utilizing American AI tools to explicitly attack the growth and infrastructure of the American AI sector itself.
According to an official investigative threat intelligence report published directly by OpenAI, the company has identified and permanently banned two distinct clusters of ChatGPT accounts executing covert foreign influence operations. Operating with clear structural indicators linking them to China, these networks systematically manipulated ChatGPT to churn out English-language social media comments, political comics, and slogans. The primary objective was to hijack real, organic American policy debates—specifically targeting public anxieties about U.S. trade tariffs and the high energy consumption of local AI data centers.
Jumping the “Data Center Bandwagon”
The first designated campaign, dubbed the “Data Center Bandwagon” by OpenAI investigators, attempted to exploit pre-existing localized frustrations over infrastructure buildouts. The threat actors used ChatGPT to pose as average American citizens on major platforms like X (formerly Twitter) and YouTube.
The accounts automatically generated commentary and visual materials alleging that the massive power grids required to sustain enterprise AI data centers were directly causing severe spikes in electricity prices for working-class American families. OpenAI’s forensic teams successfully traced this campaign node back to an unnamed, mainland Chinese technology firm that maintains multiple active software and operations contracts with regional Chinese government entities.
The second cell, labeled the “Tech and Tariffs” campaign, focused its automated output on international trade friction. These accounts generated multi-lingual political comics and text packages framing U.S. trade tariffs as a heavy-handed, unfair tactic designed to choke out global technological competition.
Crucially, OpenAI’s threat logs revealed that the originating operators explicitly commanded ChatGPT in simplified Chinese to restrict the visual scope of the output. The prompts directed the AI to exclusively focus its criticism on U.S. President Donald Trump while completely omitting any mention or depiction of Chinese President Xi Jinping, who has similarly utilized aggressive trade tariff structures.
High Hype, Minimal Traction
Despite the highly sophisticated nature of utilizing automated software to build believable personas, OpenAI noted that the actual geopolitical impact of the operation fell completely flat. The campaign registered a 1 and 2 on the Brookings breakout scale—an industry standard indicating the content made it to social media but failed to secure any meaningful engagement or virality among real human users.
“This looks like a classic example of a foreign influence operation jumping onto the bandwagon of a genuine and pre-existing domestic debate,” explained Ben Nimmo, principal investigator at OpenAI, in an interview summarized by CyberScoop. “I do want to be really clear here: this was not a case of an influence operation creating a debate. The debate existed already. We didn’t see any signs that they succeeded.”
Some of the AI-generated imagery and text displayed a stark lack of familiarity with American internet culture and linguistic nuance, resulting in clunky messaging that actual users ignored. Frustrated by their lack of traction, the “Tech and Tariffs” network even attempted a retaliatory micro-campaign targeting OpenAI itself, fabricating a completely fake narrative on X alleging that ChatGPT had suffered a massive, catastrophic user data breach.
The Future of Defensive AI
While this specific operation failed to move the needle on public opinion, the report highlights a critical trend where global adversaries are aggressively stress-testing narratives against the core foundations of Western tech dominance. The incident unfolds against a backdrop of heightened geopolitical friction; on the very same day, federal authorities announced the formal seizure of 13 internet domains linked to suspected Chinese intelligence actors using fake consulting firms to recruit U.S. military and government personnel, as reported by Reuters.
By aggressively flagging and publicizing these state-adjacent operations, tech platforms are attempting to build an immune system for digital information ecosystems, proving that the best defense against malicious AI manipulation is rigorous, real-time threat intelligence.
