OpenAI Shuts Down China-Linked ChatGPT Influence Campaigns

OpenAI says it has banned a set of China-linked accounts that were using ChatGPT to manufacture political content aimed at American audiences, with much of it pointed squarely at the AI industry itself. The company laid out the findings in a threat-intelligence report published on 10 June 2026, describing two separate operations that leaned on its chatbot to draft comments, slogans, and political cartoons.

The first operation, which OpenAI calls Data Center Bandwagon, produced material arguing that the boom in AI data centers was pushing up electricity bills for ordinary families. The second, Tech and Tariffs, generated cartoons and posts attacking US trade policy. According to OpenAI, the accounts behind the tariffs campaign gave the model oddly specific stage directions: depict President Trump, and leave Chinese President Xi Jinping out of the picture entirely.

Aimed at the AI debate

What makes the report notable is not the volume of content but its target. Both operations went after the public conversation about artificial intelligence and the infrastructure behind it. The Data Center Bandwagon material played on real anxieties about energy costs, a subject we have looked at before, and on the strain that new facilities place on local resources, which surfaced again this week in reporting on data centers rising in drought zones.

OpenAI reported that the operators used VPNs to reach ChatGPT from inside China, where the company does not allow access to its models. The content was then pushed out through what OpenAI described as a network of inauthentic accounts, much of it on X. One thread tied to the tariffs campaign falsely claimed that ChatGPT user data had been compromised, a claim OpenAI denied.

Little measurable reach

For all the effort, OpenAI found no evidence that either campaign moved public opinion. The posts drew little engagement and rarely escaped the small ring of accounts that produced them. The company has been candid about that pattern in earlier reports too: most of the covert operations it catches are clumsy and ineffective, caught well before they gain traction.

OpenAI frames the disclosure as part of its wider threat-intelligence work, sharing what it finds so that other platforms and governments can spot the same actors. The detail that lingers is the focus. State-linked operators are now spending their time trying to shape how Americans feel about the data centers and chips that power AI, treating the technology's reputation as a front worth contesting. That puts these findings alongside the recent White House order that tied AI policy to national security.

The report is also a reminder that the tools themselves are dual-use. The same model that drafts a marketing email can draft a propaganda cartoon, and the difference comes down to who is holding it.