OpenAI has banned a cluster of ChatGPT accounts tied to a Russian influence operation that manufactured pro-Kremlin content under the guise of an Israeli think tank. The operators accessed the platform from Russia through VPNs and explicitly instructed the model to scrub any linguistic fingerprints that might reveal their origin.
The campaign’s centerpiece was the “International Burke Institute,” a supposed Israel-based research outfit that published a “sovereignty index” ranking Russia favorably while making Western nations look worse. The scholarship behind it was thin. OpenAI found that 34 of 36 expert-linked articles published between September 2025 and May 2026 were lifted from other sources, some with fabricated author credits. One Cambridge University Press piece was falsely attributed to a University of Nottingham professor. A Migration Policy Institute article got credited to an Australian food chemistry professor.
Content spread across X, LinkedIn, Facebook, Substack, and Telegram, including German-language posts on a channel called “Lahme Ente” that attacked Ukraine, the EU, and Germany’s federal government while pushing for closer Moscow ties. Individual posts barely registered, but the linked Telegram channels each pulled 10,000 to 20,000 followers. OpenAI rates the operation category three on the Brookings Breakout Scale — spread across platforms with early signs of real user engagement.
This isn’t a one-off. OpenAI disrupted “Bad Grammar” in June 2024 and “Operation Helgoland Bite” a year later, both Russian campaigns using ChatGPT for political manipulation. But those are only the cases OpenAI caught on its own infrastructure. With Chinese models and open-weight alternatives freely available, the actual scale of AI-powered influence operations is almost certainly far larger than any single company’s takedown reports suggest.
