Full Report
OpenAI on Tuesday said it banned a cluster of Russian ChatGPT accounts that used VPNs to bypass access restrictions and run an influence operation, which relied on its artificial intelligence (AI) tool to generate social media posts and comments that were shared on Substack, Telegram, X, Facebook and LinkedIn. The accounts "were being used to promote the International Burke Institute (IBI), a
Analysis Summary
# Incident Report: Russian Influence Operation via International Burke Institute (IBI)
## Executive Summary
OpenAI identified and dismantled a cluster of Russian accounts using ChatGPT to facilitate a multi-platform influence operation. The campaign focused on promoting the "International Burke Institute" (IBI), a fraudulent think tank used to disseminate pro-Russian narratives and a manufactured "Sovereignty Index." While the audience reach was relatively small, the incident highlights a sophisticated shift in using AI to manufacture institutional authority and obscure the origins of state-sponsored propaganda.
## Incident Details
- **Discovery Date:** August 2026 (Reported August 26, 2026)
- **Incident Date:** Ongoing from at least February 2025 (domain registration)
- **Affected Organization:** OpenAI (Platform abuse); Social media users on X, Facebook, LinkedIn, Telegram, and Substack.
- **Sector:** Technology / Social Media / Geopolitics
- **Geography:** Russia (Origin); Global (Targeting English-speaking audiences)
## Timeline of Events
### Initial Access
- **Date/Time:** February 2025 (Initial infrastructure setup)
- **Vector:** VPN-based bypass of geographic restrictions.
- **Details:** Operators registered accounts on ChatGPT using VPNs to circumvent OpenAI's access restrictions in Russia.
### Lateral Movement
- **Details:** Not applicable in the traditional network sense; however, the operation moved laterally across social media ecosystems by creating linked profiles on Substack, Telegram, X, Facebook, and LinkedIn to amplify generated content.
### Data Exfiltration/Impact
- **Impact:** Misuse of AI models to generate social media posts, comments, and profile imagery. Creation of the "Burke Index" to project Russian military and economic self-sufficiency.
### Detection & Response
- **Detection:** Internal investigation by OpenAI triggered by suspicious patterns of AI-generated social media posts.
- **Response Actions:** OpenAI banned the identified cluster of accounts and published a report to disrupt the operation’s credibility.
## Attack Methodology
- **Initial Access:** Use of VPNs to mask Russian IP addresses and bypass regional blocks.
- **Persistence:** Maintaining multiple accounts across various social media platforms to ensure redundancy.
- **Defense Evasion:** Instructing the AI to "conceal linguistic clues" of Russian origin; using machine translation for Slavic-drafted texts to appear as native English.
- **Credential Access:** N/A (Account creation via bypassed restrictions).
- **Discovery:** Identifying and scraping academic work to misattribute and republish on the IBI website to build "authority."
- **Lateral Movement:** Cross-platform promotion of IBI content.
- **Collection:** Requesting Russian-language summaries of Telegram channel activity to monitor campaign performance.
- **Impact:** Manufacture of institutional authority through the "International Burke Institute" and its "Sovereignty Index."
## Impact Assessment
- **Financial:** Minimal direct costs reported; significant investment by actors in infrastructure and content generation.
- **Data Breach:** None; this was an abuse of service rather than a data theft.
- **Operational:** Disruption of OpenAI’s terms of service; necessitates ongoing monitoring for similar clusters.
- **Reputational:** Moderate; aims to undermine Western narratives and bolster Russian geopolitical standing through a fake "expert community."
## Indicators of Compromise
- **Network Indicators:** ibi[.]institute (Registered Feb 2025)
- **Behavioral Indicators:**
- AI-generated social media posts promoting the "Burke Index."
- Profiles using AI-generated imagery (e.g., "American Observer" Telegram channel).
- Requests to AI models specifically asking to remove Slavic linguistic markers.
- Publication of academic work with false or missing attribution.
## Response Actions
- **Containment:** Banned all identified accounts associated with the IBI cluster.
- **Eradication:** Removal of access for identified VPN-based Russian actors.
- **Recovery:** Public disclosure and collaboration with social media platforms to flag related influence content.
## Lessons Learned
- **Key Takeaways:** AI is being used less for "deepfakes" and more for "institutional fakes"—creating a veneer of credibility for fake think tanks and indices.
- **Shortcomings:** Geofencing via IP address is easily bypassed by motivated state actors using VPNs, requiring more robust behavioral detection.
## Recommendations
- **Prevention:** Enhance detection for "linguistic grooming" prompts where users attempt to hide their regional origin.
- **Monitoring:** Increase scrutiny of accounts that exhibit high-frequency posting patterns related to specific, newly registered "think tank" domains.
- **Collaboration:** Share IBI-related metadata with social media partners to ensure a cross-platform sweep of the influence operation.