“Tech and Tariffs” Campaign: Influence activity targeting US tech policy

This case study was originally published in OpenAI’s June 2026 ⁠ (opens in a new window) report on PRC-linked influence operations targeting AI debates in the US.

We banned a cluster of ChatGPT accounts that likely originated in China and used our models to generate short comments and political cartoons criticizing US tech policies and tariffs, edit work reports and assist the user in designing social media monitoring systems. The accounts prompted ChatGPT in Simplified Chinese and used VPNs to access our platform.

Their prompts repeatedly used terminology consistent with individuals associated with China’s public security system, such as seeking public opinion risk assessments concerning protests, school-bullying incidents, crowd movements in Shanghai, police-related incidents, petitioning activity and traffic enforcement. One user described the social media accounts it operated as a “water army” (水军), a common Chinese term referring to coordinated online accounts that flood platforms with criticisms or trolls, and instructed ChatGPT to generate content that would benefit the PRC or advance pro-PRC narratives. Taken together, these indicators suggest that the accounts were likely supporting activities aligned with the interests of the CCP, although we could not establish the operators’ precise institutional affiliation.

They also asked ChatGPT to propose a concept for an AI system that surveils online public opinion. They described that such a system should automatically scrape what they defined as ‘harmful’ information from ‘key persons’ on social media platforms, store logs, automatically download videos for large-scale semantic analysis and send risk notifications. Our models generated a general 500-word output that provided advice on data storage and management but did not provide ideas for how to collect that data for surveillance purposes.

Separately, we identified a network on X targeting OpenAI with false claims that ChatGPT user data had been compromised and we assessed, based on open-source behavioral indicators, that this activity was likely part of the same X network identified in this operation.

Screenshot of likely inauthentic X accounts falsely claiming ChatGPT user data was compromised.

This cluster mostly used ChatGPT to generate short comments in English and cartoon images that were posted by likely inauthentic X accounts. The most distinctive theme was US-China technological competition. The operators framed this competition around tariffs, rare earths, AI, 5G, new energy, and industrial resilience, and claimed the US was seeking technological dominance and rule-making power. They specified that the cartoons should only depict President Trump and should not include any imagery of China or China’s leader Xi Jinping.