OpenAI Reveals China-Linked Operation Using ChatGPT to Influence U.S. Policy Debate

Artificial intelligence firm OpenAI has uncovered a coordinated influence operation, suspected to be backed by China, that used its own AI tools to manipulate online discussions about AI data centers in the United States.

The operation, which OpenAI has named "Data Center Disco," employed ChatGPT to generate content for social media platforms. The prompts used to create this content were in simplified Chinese and specifically instructed the AI to focus on former U.S. President Donald Trump while avoiding any mention of Chinese President Xi Jinping. OpenAI has since blocked the accounts involved in this activity.

OpenAI identified two main types of actors involved. One group requested ChatGPT to generate graphics and commentary promoting specific viewpoints on AI, such as the argument that expanding data centers would increase electricity costs for American households. These narratives were then disseminated through social media accounts posing as ordinary Americans. The second group used AI to generate content critical of the Trump administration's tariff policies, framing them as an attempt by Trump to weaponize tariffs and maintain U.S. technological dominance.

The revelations highlight the growing concern over the use of AI by state-backed actors to influence political discourse and public opinion in other countries. The operation's focus on domestic U.S. issues, such as energy costs and trade policy, underscores the sophisticated and targeted nature of such influence campaigns.