OpenAI bans Russian cluster running fake think tank influence op

Russian operatives just got caught running one of the most elaborate AI-powered disinformation shells seen since the war in Ukraine began.

OpenAI banned a cluster of Russian ChatGPT accounts that were using VPNs to evade access restrictions. Their goal was straightforward on the surface: use Russian-language prompts to generate clean, native-sounding English posts across X, LinkedIn, Facebook, Substack, and Telegram.

Under the hood, those posts were funnels driving readers to a fake Israeli think tank called the International Burke Institute, registered back in February 2025.

Academic theft and fake metrics

IBI presented itself as a premier research hub with a physical address in Israel, claiming prominent scholars like Francis Fukuyama and Noam Chomsky as affiliated experts.

OpenAI sampled 36 research articles published under expert names on the site between September 2025 and May 2026. Exactly 34 were copied from real academic sources. The attribution was often bizarre: an authentic Cambridge University Press piece on the China-Pakistan Economic Corridor was credited to a Nottingham politics professor, while a Migration Policy Institute study was stamped with the name of an Australian food science professor.

The institute’s centerpiece was its Burke Sovereignty Index, a fabricated metric designed to elevate Russia while attacking the West. Reports slammed France under Macron as being sold off in parts, warned Germany against dependence on American bases, and framed European states as hostages of the EU.

OpenAI also spotted traces of actual individuals based in Israel representing IBI at real conferences and in article submissions, though their exact coordination with the Russian operators remains unconfirmed.

AI as an outsourced social intern

ChatGPT was not used to write the heavy research papers. Those were either lifted directly from legitimate journals or roughly translated from Slavic drafts.

Instead, operators treated ChatGPT like an offshore PR agency. They asked the chatbot to eliminate Russian linguistic markers and generate casual comments to lure Substack and social users to the think tank. One operator also used it to run a German-language Telegram channel called Lahme Ente to bash Ukraine and EU policy, while another had ChatGPT design logos for a dozen localized channels covering France, Poland, Turkey, Germany, and the U.S.

In terms of actual virality, the operation fell flat. The posts got few views, and official IBI accounts had negligible followings, though some Telegram hubs managed 10,000 to 20,000 subscribers. OpenAI categorized its impact at the lower end of Category Three on the Brookings Breakout Scale.

Why it matters

Even with modest audience numbers, this campaign shows a tactical shift worth watching closely.

Instead of spraying obvious bot spam, the operators invested heavily in laundering credibility upfront. They built a plausible institution, stole legitimate academic work, and manufactured a proprietary index that sounds serious enough for secondary citation. Generative AI wasn't tasked with hallucinating facts; it was deployed as a friction remover to translate, polish native phrasing, and handle social outreach at near-zero cost.

We think this highlights both the appeal and the tactical vulnerability of commercial AI for state-backed groups. Using ChatGPT made their outreach sound natural, but routing their operations through an AI provider's infrastructure gave OpenAI the exact breadcrumbs needed to trace and expose the broader campaign.

For platforms and researchers, text detection alone will no longer cut it. The threat is not just AI writing fake news, but AI providing operational scaffolding for fabricated institutions designed to lurk online until a geopolitical event makes them useful.