China Influence Operations Target US AI Debate, OpenAI Warns

OpenAI has disrupted two covert China influence operations that weaponized ChatGPT against America’s AI policy debate. The company detailed the takedown in its June 2026 threat report. In both cases, OpenAI banned clusters of accounts that likely originated in China.

The operators never breached any system. Instead, they used the chatbot as a content factory. They prompted ChatGPT in Simplified Chinese, relied on VPNs to bypass regional limits, and posed as everyday Americans online.

The “Data Center Bandwagon” Campaign

The first cluster pushed a pocketbook narrative. Specifically, it generated comments and images claiming that AI data center buildouts were driving up electricity prices for ordinary families.

The operators asked for comic strips and edited stock images to dramatize the message. Then likely inauthentic X accounts posted the content with hashtags such as #datacenters and #capacityauction.

OpenAI assessed that the operators likely worked for a private Chinese tech company serving provincial government clients. Notably, the same cluster also targeted overseas Chinese dissidents. When asked to insult activist Li Ying, however, “our models refused to generate inflammatory or personal attacks against Li.”

The “Tech and Tariffs” Campaign

The second China influence operations cluster took aim at trade policy. It produced cartoons attacking US tariffs as a bid for technological dominance. Strikingly, the operators instructed ChatGPT to depict only President Trump and to exclude China’s leader Xi Jinping.

This group also tried to discredit OpenAI itself. A linked network of fake X accounts spread false claims that ChatGPT user data had been compromised. OpenAI stated plainly that “these allegations were entirely false.”

One operator even described its accounts as a “water army,” a Chinese term for coordinated troll networks. The same cluster requested bulk comments designed to “benefit the PRC or advance pro-PRC narratives,” attack the US and Israel, and amplify anti-Jewish tropes.

Surveillance Ambitions and Guardrails

The actors pushed further into dangerous territory. They asked ChatGPT to design an AI system that surveils online public opinion and scrapes “harmful” information from targeted individuals.

OpenAI’s models declined to help with the surveillance collection. The system produced only general advice on data management instead. This guardrail marked an important limit on the operation’s ambitions.

Limited Reach, Larger Warning

Despite the effort, the campaigns mostly fell flat. OpenAI placed the activity at Category One on its Breakout Scale, meaning a single platform with no real spread. The firm “found no evidence of meaningful breakout beyond its own activity.”

Still, the targeting carries weight. OpenAI argued the episode shows “PRC-origin influence operators testing narratives against AI infrastructure.” You can read the full analysis in OpenAI’s June 2026 threat report.

The company drew a parallel to earlier campaigns against rare-earth firms. In those cases, PRC-linked networks tried to damage competitors in a strategic industry. Now AI appears to occupy that same role.

Why It Matters

The timing stands out. These China influence operations surfaced amid a sharp US-China trade clash and a fresh Chinese push to prioritize AI. OpenAI even noted the irony that the operators used American AI, not Chinese models, to attack American AI.

In short, the campaigns failed to move opinion. However, they preview how foreign actors may keep probing AI debates. Industry, governments, and the public should stay alert.