PRC-linked abuse: Surveillance and influence activity

This case study was originally published in OpenAI’s October 2025⁠(opens in a new window) report.

Individuals apparently linked to PRC government entities using ChatGPT to assist in development, profiling and bureaucratic functions.

As we laid out in our submission to the Office of Science and Technology Policy’s U.S. AI Action Plan in March, we believe that making sure AI benefits the most people possible means enabling AI through common-sense rules aimed at protecting people from actual harms, and building democratic AI. By democratic AI, we mean AI that is shaped by the democratic principles America has always stood for. This includes preventing the use of AI tools by authoritarian regimes to amass power and control their citizens.

Our investments in detecting and disrupting attempts by authoritarian regimes to abuse ChatGPT’s capabilities not only protect people now, but provide important insights into how those regimes might abuse future capabilities. This section discusses one set of such findings concerning individuals potentially linked to such regimes, focused on the PRC.

As we wrote in June, the PRC is making real progress in advancing its autocratic version of AI. Our disruption of ChatGPT accounts used by individuals apparently linked to Chinese government entities shines some light on the current state of AI usage in this authoritarian setting. Some elements of this usage appeared aimed at supporting large-scale monitoring of online or offline traffic, underscoring the importance of our ongoing attention to potential authoritarian abuses in this space. Other elements of this usage appeared aimed at more in-depth profiling. We banned the accounts associated with this activity and are sharing insights in this report.

Developing tools for large-scale monitoring

Profiling and research

We also banned users who appeared to be linked to Chinese government entities and who appeared to be trying to use ChatGPT for more bespoke, targeted profiling and online research.