OpenAI Busts Russian Network Using AI to Fuel Fake Think Tank

OpenAI announced on Tuesday that it banned a cluster of Russian ChatGPT accounts after uncovering a covert influence operation designed to project a favorable image of Russia and criticize Western nations. The network bypassed regional access restrictions using VPNs and leveraged AI to generate social media posts and comments across major platforms, including X, Facebook, LinkedIn, Substack, and Telegram.

The operation primarily promoted the International Burke Institute (IBI), a self-described "expert community" whose domain was registered in February 2025. Although claiming to be an Israeli-based center for global experts, IBI operated as a front. OpenAI discovered that the site featured copied academic work, misattributed articles, and texts machine-translated from Slavic languages to English. Instructed to conceal linguistic markers betraying their Russian origin, the operators aimed to construct an appearance of authority and legitimacy.

A core component of the campaign was IBI's proprietary "Sovereignty Index" (or Burke Index). Measuring nations across seven categories ranging from political and military power to cultural and cognitive sovereignty the index scored countries out of 700 points. While ranking the United States first (650.9) and China second (649.1), the report positioned Russia fourth (601.4), praising its military and economic self-sufficiency while casting its governance in a positive light.

The campaign's direct reach remained modest. Its largest audience was on Telegram, where affiliated channels like "American Observer" which used ChatGPT to generate its profile picture and summarize activity gathered between 10,000 and 20,000 followers.

OpenAI noted that the significance of this threat lies not in its audience size, but in the sophisticated infrastructure built to manufacture credibility. Rather than relying solely on automated spam, the actors integrated AI into a broader strategy involving fake think tanks, fabricated research, and pseudo-academic metrics. Ultimately, the operators' reliance on ChatGPT for promotional content provided the critical trail that exposed the entire influence network.