OpenAI Bans PRC-Linked Accounts Running Covert AI Influence Ops

OpenAI has banned two clusters of ChatGPT accounts likely originating from China after they used the company's models to run apparent covert influence operations aimed at manipulating American debates over AI and broader technology policy.

The company disclosed the operations in a report published on its website, describing a mission-driven rationale for the takedowns: identifying and disrupting "attempts by authoritarian regimes and their proxies to use AI systems to coerce critics, surveil communities or covertly interfere in democratic societies."

The first cluster, which OpenAI named the "Data Center Bandwagon" campaign, generated social media comments and images claiming that data center buildouts for AI were driving up electricity prices for average families. The second, dubbed "Tech and Tariffs," produced comments and images criticizing US tariffs as attempts to dominate technological competition.

The report includes a striking operational detail from the second campaign. The operators specified in their prompts that the content should not include China's leader Xi Jinping and should instead include only President Trump. OpenAI also linked the Tech and Tariffs cluster to a network of likely inauthentic social media accounts that were likely targeting OpenAI itself — the accounts claimed ChatGPT user data had been compromised. OpenAI states those allegations were entirely false.

Why the targeting matters

OpenAI frames the significance of these operations carefully. The targeting of OpenAI and US data center buildouts matters, the company writes, "not because the operation appears to have shifted public opinion, but because it shows PRC-origin influence operators testing narratives against AI infrastructure — a foundation of US technological leadership, economic growth and the broader democratic AI ecosystem."

In other words, the stakes are experimental. The operators appear to have been probing whether AI-generated content could be seeded into live American political fights over energy costs and the local impacts of data center development.

The campaigns did not achieve broad reach. OpenAI's assessment is direct: the operation sought to exploit and amplify existing public concerns about energy prices and local effects of data center development, but the company "found no evidence of meaningful breakout beyond its own activity."

An old playbook, a new instrument

The report situates the campaigns within a long-established pattern of foreign influence work. Foreign influence operations, OpenAI notes, have long sought to latch onto existing local issues and sincerely held beliefs, using them to build credibility, amplify divisions or exacerbate public distrust.

That is what distinguishes this operation from a crude propaganda push. The Data Center Bandwagon campaign did not invent the controversy over electricity prices — it attached itself to a genuine, ongoing argument in American communities about the costs and consequences of building out AI infrastructure. The covert element was the operator's hidden identity and motivation, not the underlying grievance.

"In this case, the operators attempted to covertly insert themselves into an ongoing American debate about the future of the country's AI capabilities while hiding who they were and what motivated them," the report states.

The tariff-focused campaign worked similarly, but with an added layer of message discipline. By instructing the model to exclude Xi Jinping and feature only Trump, the operators shaped generated content to direct criticism at one government while shielding another — a targeted framing choice embedded directly in the prompts.

The false claims about ChatGPT user data add a second dimension to the operation: an apparent attempt to undermine trust in OpenAI specifically, not just to influence policy debate. Combining narrative laundering with reputational attacks against the target company suggests operators thinking about AI firms as both tools and adversaries.

Democratic AI as a stated mission

OpenAI opens the report by positioning the takedowns within a broader ideological framework. The company describes its mission as ensuring that artificial general intelligence benefits all of humanity, advanced by "deploying our innovations to build democratic AI: AI shaped by democratic principles, governed by common-sense rules and designed to help people solve hard problems while protecting them from real harm."

That mission, the report argues, requires countering authoritarian use of the same technology. OpenAI characterizes the adversary as "totalitarianism with AI characteristics — the use of AI for surveillance, censorship and control over political, social and private life."

The publication itself is framed as a defensive measure. By releasing the findings, OpenAI says it aims to help industry, governments, civil society and the public better identify and disrupt attempts by foreign threat actors to manipulate legitimate public debates, weaken democratic institutions and advance authoritarian aims with AI.

The context for readers

The disclosure lands amid an intensifying US debate over AI infrastructure. Data center construction has become a flashpoint in state and local politics, with communities and regulators weighing energy demands, grid costs and environmental impacts against the economic pull of AI investment. That debate is genuine and contested. It is also, as this report demonstrates, now a target for foreign influence operators who see an opportunity to deepen existing divisions without revealing their hand.

The tariff angle connects the operation to another live policy fight. US tariff policy toward China has become central to the technological competition between the two countries, and the Tech and Tariffs campaign sought to frame those measures as aggression rather than strategy — again, by covertly inserting generated content into a debate Americans are already having.

The report also signals how AI vendors increasingly see themselves as frontline actors in information integrity. OpenAI did not merely remove accounts; it published campaign names, tactics and attributions, treating transparency about influence operations as part of its product responsibility.

What the operations lacked in reach, they made up for in specificity. The operators chose targets at the intersection of economic anxiety and geopolitical competition, used prompt-level instructions to control political framing, and paired policy messaging with disinformation about the AI company whose tools they were exploiting.

OpenAI's assessment suggests these were tests rather than successes — but tests establish capability and intent. As AI infrastructure becomes more central to US economic and political life, and as generative tools lower the cost of producing persuasive content at scale, the expectation embedded in OpenAI's report is that similar campaigns will return, targeting the same debates with refined methods.