A supposed expert institute publishing academic-looking research was actually part of a covert influence operation supported by ChatGPT, according to OpenAI.
The company said Tuesday it banned a cluster of accounts originating in Russia that were used to promote the International Burke Institute, a self-described Israel-based expert community. OpenAI’s investigation uncovered copied academic research, misleading author attributions and a “sovereignty index” that portrayed Russia favorably while criticizing Western governments.
The operation shows how generative AI can support more than mass-produced social posts. It can also help operators build the surrounding infrastructure — research, branding and coordinated accounts — needed to make an influence campaign look more legitimate.
A fake institution with real-looking research
The operation centered on the International Burke Institute, whose website was registered in February 2025 and presented itself as an Israel-based “expert community.”
OpenAI reviewed 36 articles published on the site between September 2025 and May 2026 and found that 34 appeared to have been copied from elsewhere, sometimes with incorrect authors. One article on the China-Pakistan Economic Corridor, for example, appeared to originate from Cambridge University Press but was attributed to a different academic.
The site also claimed connections to prominent figures including Francis Fukuyama and Noam Chomsky, according to OpenAI.
The institute’s most distinctive feature was its “sovereignty index,” which rated countries across political, economic, technological, information, cultural, cognitive and military measures. OpenAI said the index was used to praise Russia and attack Western governments.
Must-read security coverage
- UK Police Convicts Pair in £5.5 Billion Bitcoin Launder Case
- Blackpoint Cyber vs. Arctic Wolf: Which MDR Solution is Right for You?
- How GitHub Is Securing the Software Supply Chain
- 8 Best Enterprise Password Managers
AI was the amplifier, not the whole operation
The ChatGPT accounts generated promotional posts and replies designed to push people toward IBI content. One operator also produced German-language material for a Telegram channel called “Lahme Ente,” which criticized Ukraine, the EU and Germany while advocating closer relations with Russia.
Another created logos for about a dozen Telegram channels focused on countries including Germany, the U.S., France, Poland and Türkiye and used ChatGPT to summarize their activity in Russian.
OpenAI said the campaign’s immediate reach was modest. Most posts received few views, while the associated Telegram channels generally had 10,000 to 20,000 followers each.
That limited reach does not make the operation insignificant. The more important development is the way AI was used as one component of a larger credibility-building system. OpenAI said, “The significance of the operation lies less in the audience it reached, however, than in the infrastructure it had built.”
That distinction matters for platforms, businesses and consumers. Generating a convincing post is increasingly easy; building an apparently independent institution around that content could also make coordinated messaging look more credible and therefore harder for users to recognize as part of an influence operation.
OpenAI said activity involving ChatGPT also gave investigators signals that helped connect parts of the wider campaign. That creates an unusual dynamic for influence operators: the same AI tools that make coordinated activity easier to scale can also leave patterns that platforms can use to uncover it.
More news: Microsoft Teams is rolling out a new policy that lets IT administrators automatically block detected external meeting bots, giving organizations more control over third-party AI assistants in sensitive calls.