OpenAI Bans Russia-Linked ChatGPT Accounts Behind Fake 'Think Tank' Influence Campaign
OpenAI has disrupted a covert influence operation, very likely originating in Russia, that used ChatGPT to build and promote a fake Israel-based think tank called the International Burke Institute (IBI). The operators banned by OpenAI used the accounts to generate articles, social media posts, and a fabricated “sovereignty index” designed to cast Russia favorably while attacking Western governments.
Details
- Attribution: OpenAI identified “a cluster of ChatGPT accounts that very likely originated in Russia.” Operators routed traffic through VPNs to bypass geographic restrictions, prompted ChatGPT in Russian, and explicitly instructed the model to hide linguistic clues that would reveal their Russian origin
- The fake think tank: The operation built a website for the “International Burke Institute,” registered in February 2025, which falsely claimed to be based in Israel and listed purported experts — including names like Francis Fukuyama and Noam Chomsky — who had no actual involvement
- The “sovereignty index”: The IBI published a proprietary “sovereignty index,” a metric engineered to portray Russia favorably while criticizing Western nations. Sample articles attacked France (claiming its “historical capital” was “being sold off in parts”) and Germany (arguing it needed “real military sovereignty, not dependence on American bases”)
- Plagiarized content: Of 36 sampled IBI articles published between September 2025 and May 2026, OpenAI found 34 had been copied from elsewhere on the internet, often republished under false attribution to manufacture an appearance of original scholarship
- How ChatGPT was used: The accounts used ChatGPT to draft social media comments and promotional posts across X, LinkedIn, Facebook, Substack, and Telegram, write German-language posts criticizing Ukraine, the EU, and Germany, design logos for about a dozen Telegram channels, and reply to real users urging them to follow the IBI’s channels
- Targeting: Beyond the main IBI presence, the network ran custom Telegram channels focused on Germany, the United States, France, Poland, and Turkey
- Reach: OpenAI assessed the operation’s actual impact as limited — posts typically drew low view counts, and using the Brookings Breakout Scale, the campaign ranked at the low end of Category Three (multi-platform activity with only modest genuine audience penetration). The Telegram channels had accumulated roughly 10,000–20,000 followers each
- OpenAI’s assessment: The company said the case illustrates how influence actors use AI “as a supporting tool within a broader effort to manufacture authority, obscure the source of favored narratives, and establish assets that could be scaled over time” — emphasizing that AI functions as an amplifier within a larger disinformation toolkit rather than a novel threat on its own
What happened next
- OpenAI banned the identified ChatGPT accounts as part of its ongoing enforcement against coordinated inauthentic behavior and influence operations
- The company published the findings publicly as part of its recurring threat-intelligence reporting on misuse of its models
- Read OpenAI’s full report at https://openai.com/index/disrupting-malicious-uses-of-ai-influence-campaign-russia