Full Breakdown
OpenAI Disrupts Russian-Run AI Influence Network Centered on a Fake Think Tank
8/26/2026, 8:58:54 PM
Operation Unveiled: Ban of Russian ChatGPT Accounts
On August 25, 2026, OpenAI disabled a cluster of ChatGPT accounts traced to operators in Russia. The accounts accessed the service via VPNs to bypass OpenAI’s geo-restriction, which blocks model access from Russian IP addresses. OpenAI’s Intelligence & Investigations team said the accounts were part of a covert influence campaign that used the chatbot to generate English-language social-media posts promoting the International Burke Institute (IBI), a self-described “expert community” that claimed an Israeli base.
Background & Context
The IBI website, registered in February 2025, presented itself as a global research institute listing scholars such as Francis Fukuyama and Noam Chomsky. A review of 36 articles attributed to the institute’s “experts” found that 34 were copied from existing works, many with false bylines or machine-translated from Slavic languages. The site also hosted a proprietary “sovereignty index” that awarded Russia a score of 601.4 while giving lower scores to the United States (650.9), China (649.1), Switzerland (610.7) and several Western nations.
Timeline
- February 2025 – Domain for the International Burke Institute registered.
- September 2025 – May 2026 – Period during which the 36 sampled IBI articles were published.
- July 17 – The IBI site listed top-tier names, including Fukuyama and Chomsky, on its roster.
- August 25 2026 – OpenAI bans the Russian-origin ChatGPT accounts after an internal investigation.
Data & Statistics
- The operation posted on Substack, Telegram, X (formerly Twitter), Facebook and LinkedIn.
- Telegram channels linked to the campaign attracted between 10,000 and 20,000 followers each.
- OpenAI placed the campaign at the lower end of Category 3 on the Brookings Breakout Scale.
- Of the 36 IBI articles examined, 34 were plagiarized or misattributed; none were generated by ChatGPT.
- The sovereignty index scores: United States 650.9, China 649.1, Switzerland 610.7, Russia 601.4.
Official Statements & Responses
OpenAI’s blog post framed the discovery as an escalation of its threat-reporting program, noting that the investigation began with “AI-generated social-media posts” and expanded to reveal a broader network that combined AI-assisted content creation with a fabricated institutional front. The company emphasized that the operation’s significance lay in the “infrastructure it had built” rather than the modest audience reach of individual posts. Operators deliberately instructed ChatGPT to strip linguistic cues that could reveal a Russian origin, leaving detectable translation artifacts such as the literal rendering of the German “Ampel-Koalition” as “Svetofor coalition.”
Verbatim Quotes
- “The significance of the operation lies less in the audience it reached, however, than in the infrastructure it had built,” — Noam Chomsky, OpenAI.
Why It Matters
OpenAI’s disclosure shows how generative AI can serve as a low-cost tool for influence actors seeking credibility. By coupling AI-produced promotional posts with a seemingly scholarly website, the operators created a veneer of authority that could be scaled if the infrastructure were expanded. The case also demonstrates that AI-driven campaigns may be detectable through patterns in prompt language, VPN usage, and anomalous translation errors, offering a potential avenue for platforms and investigators to intervene early.
What’s Next
OpenAI indicated its threat-intelligence team will continue monitoring for similar misuse of its models. The company has disrupted more than 40 malicious networks since early 2024, including a China-linked “Data Center Bandwagon” campaign in June 2026. No further enforcement actions related to the IBI network have been announced.
