Openai bans China-linked accounts over fake electricity bill outrage

OpenAI has banned two clusters of ChatGPT accounts it believes were operating from China after finding they used the platform to run covert influence campaigns aimed at US tech and policy debates, including one effort that produced cartoons blaming AI data centres for rising household electricity bills.

The first cluster, dubbed "Data Center Bandwagon", generated social media comments and comic strips tying AI data centres to higher electricity costs. According to OpenAI, operators prompted ChatGPT in Simplified Chinese via VPNs while posing as Americans from varied backgrounds on X.

The company assessed the group was likely a social media team at a private Chinese tech company working on behalf of provincial government clients. Despite the coordinated effort, OpenAI rated the activity Category One on its Breakout Scale, meaning it stayed confined to one platform with no evidence it reached real audiences.

"This was not a case of an influence operation creating a debate," said Ben Nimmo, OpenAI's principal investigator.

However, while there was no particular grid operator named in the cartoons regarding auction pricing of capacities, it is an issue that has been sufficiently discussed. In relation to data centres, the independent market monitor of the PJM Interconnection indicated that their high demand has led to a rapid increase in power prices in the region.

Furthermore, three United States senators have also individually asked Amazon, Google, and Meta to clarify how increased prices have affected their residential customers, showing that a propaganda effort overlapped an ongoing policy debate.

The other network called "Tech and Tariffs" created cartoons against tariffs in a manner that did not depict Xi Jinping but instead featured bulk comments in English, Italian, Japanese, and traditional Chinese, which was intended for audiences in Taiwan.

The owner of one account referred to this network as a “water army” and requested ChatGPT's assistance in developing technology to scrape and detect trolls. OpenAI noted that ChatGPT declined and simply provided some generic information.

Accounts from this same network spread misinformation regarding ChatGPT and its database being hacked, which, according to OpenAI, was an effort to discredit ChatGPT’s reputation.