OpenAI bans Russian network that built a fake academic institute to spread pro-Moscow content

OpenAI banned a cluster of ChatGPT accounts on Tuesday that powered a Russia-linked influence operation built around a fake think tank called the International Burke Institute and a branded 'sovereignty index' that ranked countries to Russia's advantage, the company said in a blog post. The posts themselves reached a small audience, OpenAI wrote. The apparatus around them was built to be reused.

The campaign's spine was the IBI, a fictitious policy organization whose website mixed recycled academic writing, sometimes with false attribution, with original commentary that framed Russian interests in the language of Western foreign-policy debate. Beside it sat the sovereignty index, a country-ranking product built to look like a respectable research output. Together, the two assets gave the operators something a bare social-media account never could: the surface of authority, with the cost of generating it collapsed to near zero by ChatGPT.

According to OpenAI, the operators prompted ChatGPT in Russian to draft posts in English, then asked the model to scrub any linguistic signal that would betray a non-native author. The finished comments were pushed to Substack, Telegram, X, Facebook, and LinkedIn, the five platforms CNBC named in its write-up of the takedown. OpenAI does not sell access to its models inside Russia, and the operators used VPNs to obscure their location, CNBC reported.

What the operation lacked in reach it compensated for in structure. OpenAI's own framing in the blog post is that the significance lies less in the audience it reached than in the infrastructure it had built. The IBI website, the recycled academic cache, the branded index, and the language-hiding prompt template are each reusable parts. A second operator, or the same one at a different domain, could drop in a new set of papers, a new ranking, and a new platform mix and be operational in days rather than months.

In February 2025 OpenAI disrupted accounts linked to Rybar, a pro-Russia media organization the UK government has described as partially coordinated by the Russian Presidential Administration, funded by the state defense conglomerate Rostec, and staffed by people who have worked with Russian intelligence services. Time has previously reported that Russia, China, and Israel have all used OpenAI's tools in foreign-influence work, and The Hacker News has documented Russian, North Korean, and Chinese actors misusing ChatGPT for cyber operations.

The second sighting in eighteen months is what makes the new takedown more than a press release. The IBI-plus-sovereignty-index template is now a known shape. Threat-intel teams at other platforms, the research desks of Western think tanks, and election-integrity units in the EU and U.S. can scan for the same fingerprints: a policy-sounding institute with no institutional history, an index that conveniently favors one country, recycled or misattributed academic citations, and posts that read like fluent native commentary but were drafted in another language through a chatbot.

There are limits to how much weight to put on this one case. OpenAI does not publish account counts, post counts, or follower reach for the campaign, and the CNBC report does not add them. Russia's embassy in London and the Foreign Ministry in Moscow had not commented at the time of OpenAI's announcement, so the operation's backers remain a characterization rather than a confirmed attribution. The UK government's description of Rybar was relayed by OpenAI and CNBC rather than independently re-verified here.

What is verifiable is the artifact class. The next operation does not have to be Russian, and it does not have to use OpenAI. But if it shows up with a fresh institute, a recycled-paper library, a convenient country ranking, and a translation pass to scrub the language, the defenders will at least know what they are looking at before the posts start to spread.