OpenAI Bans a Russia-Origin Cluster Promoting a Fake Israel Think Tank
On August 25, 2026, OpenAI’s Global Affairs team said it banned a cluster of Russia-origin ChatGPT accounts used to promote the International Burke Institute, a self-described Israel-based “expert community” whose site copied academic work, ran a sovereignty index praising Russia, and hid Slavic authorship. Impact was limited (Brookings Breakout Scale Category 3, low end); the tell is the infrastructure.
TLDR
OpenAI on Tuesday, August 25, 2026 (Global Affairs) published “Disrupting a new covert influence campaign from Russia.” It banned a cluster of ChatGPT accounts that very likely originated in Russia (operators prompted in Russian, used VPNs because Russia is blocked). The cluster generated English social posts promoting the International Burke Institute (IBI)—a site registered February 2025 that claimed an Israel address and a roster including Fukuyama and Chomsky. Of 36 sampled articles (Sep 2025–May 2026), 34 were copied from elsewhere, often misattributed. IBI’s sovereignty index praised Russia and attacked France, Germany, the EU, and the US. ChatGPT use: promotional posts and replies on X, LinkedIn, Facebook, Substack, Telegram; German-language posts on “Lahme Ente”; logos for a dozen Telegram channels. OpenAI rates impact at the low end of Brookings Breakout Scale Category 3. Telegram channels: typically 10–20,000 followers.
What OpenAI says it found
| Item | openai.com Global Affairs Aug 25 |
|---|---|
| Action | Banned Russia-origin ChatGPT cluster |
| Front | International Burke Institute — claimed Israel; not ChatGPT-written core essays |
| AI role | Social promotion, replies, Telegram logos, Russian-language summaries |
| Giveaway | “Svetofor coalition” (Slavic calque of Germany’s traffic-light coalition) |
| Copied corpus | 34 / 36 sampled articles lifted; wrong authors (e.g. CUP paper attributed to the wrong professor) |
| Audience | Low views on social; Telegram 10–20k; Category 3 low end |
| Caveat | A handful of real people in Israel may have represented IBI; OpenAI cannot determine the relationship |
This is not a model GA and not a claim that ChatGPT wrote the sovereignty reports. The lab’s point: AI as a supporting tool inside a broader authority-laundering operation, and the AI use is what made the cluster detectable.
Product-line de-dupe: not prior OpenAI threat-intel PDFs (2024–2025 series cited in the post), not Aug 7 Astra Critical, not Aug 18 pacing, not Aug 24 Alabama subpoena. Influence-ops disruption is a separate Global Affairs lane.
Why this story matters
The interesting failure mode is not viral deepfakes. It is a plausible institute, a proprietary index, and stolen papers—with ChatGPT used only to spam the distribution layer. That is cheap to copy and hard for platforms to label as state media. Watch: whether IBI stays up, whether Israel-based names are identified, and whether the next report shows the same operators moving to open-weight models after the ban.