OpenAI says Chinese-linked propagandists used ChatGPT to target Trump policies

AI tools built to speed work and creativity can just as easily be turned toward America’s most combustible policy fights. OpenAI said Chinese-linked propagandists used ChatGPT to generate anti-Trump slogans, cartoons and social posts aimed at tariffs and data centers, exposing how quickly generative AI can be repurposed to pressure public debate in the United States.

OpenAI said it identified and banned two clusters of accounts it believed likely originated in China and were built to shape arguments over President Donald Trump’s trade and technology agenda. The company named the operations Data Center Bandwagon and Tech and Tariffs. One cluster generated social media comments and images claiming AI data center buildouts were driving up electricity prices for average families, a message designed to exploit fears over household bills and local development. The other produced comments and images attacking U.S. tariffs as an attempt to dominate technological competition.

The tariff-related effort was especially specific. OpenAI said the prompts were written in Simplified Chinese and instructed the model not to include Xi Jinping, only Trump. Some of the resulting images were later posted on X. OpenAI described cartoons showing Trump behaving disruptively on the world stage, including scenes meant to suggest he was breaking the global order or undermining economic stability. Another group generated Chinese comments for Chinese-language articles, along with content in Italian and Japanese, underscoring that the campaign was not confined to a single language or audience.

OpenAI said one cluster appeared tied to a Chinese tech company that did government work, though it did not identify the firm. The company said the activity dated to late 2025 and early 2026 and found no evidence of meaningful breakout or significant public-opinion shift. Ben Nimmo, OpenAI’s principal investigator, said the operations appeared aimed at manipulating a legitimate debate about American tech policy, highlighting the irony that American AI tools were being used in the campaign.

The findings land in a U.S. political fight that is already volatile. OpenAI said the anti-data-center content tried to tap public concern over electricity prices and the local impacts of construction. POLITICO has described the issue as a policy flashpoint heading into the midterms, and the National Conference of State Legislatures said lawmakers in 14 states were considering banning data center development as of June 2, 2026. The Chinese Embassy in Washington denied the accusations, saying it opposed groundless attacks or smears against China and wanted AI to be a force for good.