Two clusters of ChatGPT accounts, likely originating from China, have been banned after OpenAI caught them running covert influence campaigns targeting US AI and trade policy debates — and one of them even tried to frame OpenAI for a fake data breach.
Two Campaigns, One Playbook
The first cluster, which OpenAI named “Data Center Bandwagon,” pumped out social media comments and images claiming that AI data center buildouts were driving up electricity prices for average families. Not a sophisticated psy-op — just latching onto a real American concern about energy costs and trying to twist it into opposition against AI infrastructure.
The second cluster, “Tech and Tariffs,” generated comments attacking US tariffs as a tool for dominating tech competition. The operators explicitly instructed ChatGPT to avoid any mention of China’s leader Xi Jinping, while only referencing President Trump. That same network also published false allegations that OpenAI had suffered a user data compromise — a smear designed to undermine trust in the company itself.
Why This Matters More Than the Reach
OpenAI found no evidence that either operation broke out of its own bubble or shifted any real public opinion. The significance isn’t scale — it’s target selection. These operators chose to test narratives against US data center buildouts, which are the physical backbone of American AI leadership. That’s a shift from generic political meddling toward attacking the infrastructure of a strategic industry.
Foreign influence operations have always exploited existing local anxieties. What’s new here is the weaponized use of generative AI to produce high-volume, on-message content with specific narrative constraints (e.g., “don’t mention Xi”). The bans happened after OpenAI’s internal detection systems flagged the activity, and the company published the findings to help the industry and governments recognize similar patterns.
What This Enables Next
Expect more targeted, AI-generated influence campaigns that probe critical infrastructure debates — energy, supply chains, semiconductor fab locations, and data center siting. The playbook is now public, but the cat-and-mouse game between detection and evasion just got a lot harder to track at scale.
Source: PRC-linked influence operations are targeting AI debates in the US
Domain: openai.com
Comments load interactively on the live page.