Those “Data Centers Are Raising Your Electricity Bill” Posts? They Were ChatGPT-Powered Propaganda!

Hi, it's Shiichan! I found a news piece today that gave me a little shiver — it turns out some of those "data centers are raising your electricity bill" social posts were actually AI-generated propaganda. Let me walk you through it!

OpenAI News openai.com

What was announced?

OpenAI's News published a case study, originally part of a June 2026 threat report, revealing that it banned a cluster of ChatGPT accounts likely originating in China. The cluster used ChatGPT to mass-produce social media content — comments and images — criticizing US data centers and AI infrastructure. OpenAI believes the operators were likely part of a social media operations team at a private Chinese tech company doing work for Chinese provincial government clients.

Why it matters

OpenAI normally blocks access from China, but this cluster got around that using VPNs. They prompted ChatGPT in Simplified Chinese while requesting both English and Chinese outputs, all while posing as Americans. OpenAI's previous threat report already flagged harassment against Chinese dissident Li Ying, and this new case shows an even more elaborate playbook — a clear sign that AI-powered, cross-border influence operations keep getting more sophisticated.

What changes

With these accounts banned, at least this particular channel of posts has been shut down. By publishing the tactics, OpenAI is also giving readers a reason to pause the next time they see a "data centers are raising your bill" post online — is it a genuine grassroots complaint, or an organized influence campaign? Platforms like X and Facebook can also use these specifics to sharpen their own fake-account detection.

Dive Deep

The cluster's main focus was steering opinion around US electricity infrastructure. Drawing on real reporting from a regional grid operator, they had ChatGPT create comic strips about rising capacity auction prices, framing the extra demand as coming from data centers and AI, with the cost ultimately passed on to ordinary households.

• Posted on X with hashtags like #capacityauction, #datacentersuccess, and #datacenters, alongside links to legitimate news about the grid operator's capacity auctions
• Edited generic electricity-market images by adding text, reinforcing a narrative that ordinary people were subsidizing AI infrastructure
• Had ChatGPT write code to automate logins and manage accounts across multiple platforms
• Used ChatGPT as a text-processing tool — extracting usernames, formatting links, and preparing worksheets

The cluster's second target was the overseas Chinese community. They tried to get ChatGPT to write insulting comments about Chinese dissident Li Ying (also known as "Teacher Li") and other commentators like Lu Yiheng, Xu Chi, and the account @SydneyDaddy1 — though the model refused to generate personal attacks against Li. In an even more novel move, the operators posed as US-based Chinese immigrants, workers, and students to encourage a US-based former Chinese police officer to criticize US policy failures on a YouTuber's channel.

Work reports the cluster had ChatGPT polish also revealed their strategy for evading platform detection while keeping accounts credible long-term. One report focused specifically on Facebook, describing how to build trust through everyday-life posts, coordinate multiple accounts to make engagement look organic, combine organic posting with paid ads, and keep backup accounts ready to spread operational risk.

Wrap-up

• OpenAI banned a cluster of ChatGPT accounts likely originating in China
• The cluster mass-produced social posts and images criticizing US data centers and AI infrastructure
• It accessed ChatGPT via VPN, prompted in Simplified Chinese, and posed as Americans
• It also targeted overseas Chinese dissidents with harassment and impersonation-based influence tactics
• ChatGPT was also used to automate logins and account management for the operation

If you're curious about how AI misuse and platform influence operations actually work, this one's worth a read — and it's a good reminder to pause the next time a "data center backlash" post crosses your feed.