That "Think Tank" ChatGPT Helped Build? It Was a Russian Influence Operation!

Hi, it's Shiichan! Today I want to walk you through a pretty elaborate Russia-linked influence operation that OpenAI just exposed.

OpenAI News openai.com

What was announced?

OpenAI's News page announced that it banned a cluster of ChatGPT accounts that very likely originated in Russia. These accounts were being used to generate social media posts promoting the International Burke Institute (IBI), a self-described "expert community" claiming to be based in Israel. What started as an investigation into AI-generated social media posts grew into a much bigger operation: a website padded with plagiarized and misattributed academic work, a homegrown "sovereignty index" that praised Russia while criticizing the West, and efforts to hide the operators' Russian origins.

Why it matters

OpenAI has disrupted several Russia-linked operations since the start of the war in Ukraine, but this one stands out for how elaborate it was. Rather than just seeding posts across social media, the operators built out a whole credible-looking expert institution, complete with plagiarized and misattributed academic papers, to make it look authoritative. It's the first time OpenAI has disrupted a Russia-tied operation that went to such lengths, and it was a previously unreported campaign.

What changes

OpenAI banned the accounts involved and published a full writeup of how the operation worked. The campaign itself doesn't appear to have spread very widely, but exposing this playbook, pairing fake credentials, plagiarized-but-authentic-looking papers, and a proprietary index, gives everyone a clearer picture of how this kind of fake authority gets manufactured. Even when the AI-generated part is just a handful of promotional posts, the fabricated "think tank" behind it is the real risk if people take it at face value.

Dive Deep

The operators prompted ChatGPT in Russian to generate English-language social media comments for Substack, Telegram, X, Facebook, and LinkedIn. Since OpenAI doesn't allow access from Russia, they used VPNs to reach the platform, and there's evidence they explicitly instructed ChatGPT to strip out any linguistic clues pointing to their Russian origin.

The IBI website was registered in February 2025 and claimed to be based in Israel. The articles on it weren't AI-generated. Many were copied wholesale from real academic writing and attributed to the wrong people. Of 36 articles OpenAI reviewed that were published between September 2025 and May 2026 and linked to IBI's listed experts, 34 turned out to be copied from elsewhere. One example: an article on the China-Pakistan Economic Corridor was lifted from a Cambridge University Press original but credited to a different professor at the University of Nottingham whose actual expertise is South Asian politics. Another, on migration governance, was copied from the Migration Policy Institute but attributed to an Australian food science professor instead.

The site's centerpiece was its "sovereignty index," with reports rating countries like France, Germany, the EU, and the US, mostly ones that have criticized Russia's war on Ukraine, in sharply critical, polemical terms.

On the AI side, the main use of ChatGPT was generating social posts promoting IBI's articles, posted both by accounts bearing the IBI name and by accounts posing as ordinary users. A separate operator also generated German-language posts for a Telegram channel called "Lahme Ente" ("lame duck"), which criticized Ukraine, the EU, and the German government while pushing for closer ties with Russia. A second operator generated logos for a dozen Telegram channels covering Germany, the US, France, Poland, and Türkiye, including one called "American Observer" posing as a US outlet, whose bio contained oddly non-native English phrasing that reads like a machine translation.

The overall reach was limited: view counts on the posts and follower counts on the official IBI accounts were both low. The Telegram channels did somewhat better, generally landing in the 10,000-20,000 follower range. Using the Brookings Breakout Scale, OpenAI rated the operation at the lower end of Category Three, meaning it reached multiple platforms with some signs of breaking out to authentic audiences.

Wrap-up

• OpenAI banned a cluster of Russia-based ChatGPT accounts promoting IBI, a fake Israel-based "expert community"
• The IBI website plagiarized and misattributed real academic papers, and pushed a "sovereignty index" praising Russia and criticizing Western countries
• Operators prompted in Russian via VPN, generating English social posts and German Telegram content, while trying to hide their Russian origins
• Overall reach was limited, landing at the lower end of Breakout Scale Category Three
• Worth a read whether you follow AI safety and disinformation research, or you just want to be a little more skeptical the next time you see an "expert institute" cited online.