OpenAI disrupts a Russian influence campaign posing as a think tank

OpenAI said it had banned a cluster of ChatGPT accounts, operated from Russia via VPNs to evade the company’s block on Russian access, that had used the model to build and promote the International Burke Institute — a self-described think tank, registered in Israel in February 2025, that published a “Sovereignty Index” ranking countries favourably toward Russia and against Western support for Ukraine.

OpenAI’s investigation found the institute’s credibility was largely fabricated: of 36 articles attributed to its listed experts, 34 had in fact been copied from elsewhere and misattributed, in some cases to real academics — reportedly including Noam Chomsky and Francis Fukuyama — some of whom had died before the institute existed. The operators used ChatGPT to draft social media comments in Russian and to strip out linguistic markers that might reveal their origin, then posted the results to Substack, Telegram, X, Facebook and LinkedIn.

OpenAI assessed the campaign’s actual reach as limited — most individual posts and the institute’s own accounts drew few views or subscribers — though it noted some of the operation’s Telegram channels had accumulated between 10,000 and 20,000 followers each. The disclosure was the latest in a recurring series of threat-intelligence reports OpenAI has published since 2024 on state-linked misuse of its models, previously tied to Russia, China, Iran and North Korea.