OpenAI has banned a cluster of ChatGPT accounts it says were part of a pro-Russia influence operation built around the International Burke Institute, a self-described 'expert community' based in Israel that OpenAI identifies as the 'core' of the campaign.
According to OpenAI, the banned users relied on VPNs to evade the company's restrictions on Russian users, then used ChatGPT to generate social media posts bashing Western countries and promoting the institute's work, including a 'sovereignty index' that favored Russia.
The numbers come first. OpenAI's review found that 34 of 36 articles published on the Burke Institute's website were real academic work misattributed to other authors. One example: a legitimate migration-governance article from the Migration Policy Institute was falsely credited to Kate Howell, a University of Melbourne food-science professor, complete with a headshot pulled from the internet. After OpenAI's blog post, the site quietly corrected the byline to Kate Hooper, the actual author.
The institute's contributor list also named Noam Chomsky and Francis Fukuyama, along with experts who died before the institute was even founded, such as Shlomo Avineri and Jiang Ping. Neither Chomsky nor Fukuyama responded to Fortune's requests for comment. Susan Ariel Aaronson, a George Washington University professor whose article also appeared on the site without her knowledge, told Fortune the situation was 'disturbing' and said trust in AI governance depends on catching exactly this kind of fraud.
OpenAI said the website was registered in February 2025 but that nothing was published on it before October 2025, per a Fortune review using the Wayback Machine. Individual posts promoting the institute drew few views, OpenAI said, but Telegram channels tied to the operation attracted between 10,000 and 20,000 followers apiece.
The Burke Institute pushed back through French outlet Le Monde, arguing that third parties citing its publications or its Sovereignty Index 'does not, in itself, constitute evidence of cooperation, coordination, funding, or a common underlying purpose.' The institute did not respond to Fortune's direct request for comment.
Aaronson also argued that federal regulators are doing little to police this kind of activity, calling the current approach opaque. That complaint deserves scrutiny rather than automatic deference.
Capital rewards clear rules, and reputational risk did more work here than any statute on the books. OpenAI moved to protect its platform and its brand — not because Washington ordered it to, but because a company with earnings, margins, and users to answer to has every incentive to keep its product free of foreign manipulation. The market has already voted: private enforcement caught what a slow-moving administrative state did not.
Calls for heavier federal oversight of artificial intelligence should reckon with this episode rather than skip past it. The fraud was identified, documented, and dismantled by the very firm critics say needs more supervision. That is the case for free enterprise policing itself, not for handing Washington a bigger rulebook. Power, as ever, leaves a paper trail — and in this case, it was a private company that read it first.
