Loading page…
The operation had limited direct reach, but OpenAI says its institutional-looking website, purported experts and geopolitical index created infrastructure that could be expanded over time.
Listen to this story
OpenAI’s investigation uncovered a Russia-origin influence network using ChatGPT through VPNs to produce multilingual political content and build credibility for the purported International Burke Institute. The institute published 34 copied or misattributed articles out of 36 reviewed between September 2025 and May 2026, alongside a Russia-favorable “sovereignty index.” Distribution reached Substack, Telegram, X, Facebook and LinkedIn, but OpenAI rated the campaign only 3 of 6 for audience breakout.
Operators used VPNs because OpenAI does not provide ChatGPT access in Russia.
The operation generated Russian- and German-language material targeting Ukraine, Germany, the EU and Western governments.
One operator created branding for roughly 12 Telegram channels spanning Germany, the U.S., France, Poland and Turkey.
OpenAI has banned a cluster of Russia-origin ChatGPT accounts, shutting down a content-generation route used by a covert influence campaign. The company said it traced AI-generated social posts to a broader operation built to make pro-Russia narratives look independently sourced.
The campaign promoted the International Burke Institute, a purported Israel-based think tank. Its website featured copied and misattributed academic work, while a “sovereignty index” portrayed Russia favorably and denigrated Western countries.
OpenAI said ChatGPT-produced posts directed readers to an otherwise credible-appearing institution, complete with purported experts, republished academic work and a purported proprietary risk index. The company said that structure could help manufacture authority and obscure the source of favored narratives.
OpenAI found that 34 of 36 expert-linked institute articles published between September 2025 and May 2026 had been copied from other sources. Some used false author credits.
The operators used virtual private networks, or VPNs, to disguise their location because OpenAI does not allow model access in Russia. They prompted ChatGPT in Russian to generate social posts and remove linguistic clues that could reveal their origin.
OpenAI characterized the model as a supporting tool within a wider operation, rather than the operation’s sole asset. Its investigation of AI-generated posts was what led the company to the broader network.
Generated material appeared on Substack, Telegram, X, Facebook and LinkedIn. The campaign also included German-language posts criticizing Ukraine, the European Union and Germany’s federal government while promoting closer ties with Russia.
One operator created logos for about a dozen Telegram channels focused on Germany, the U.S., France, Poland and Turkey, and regularly requested Russian-language summaries of channel activity.
OpenAI has previously reported disrupting Bad Grammar in June 2024 and Operation Helgoland Bite in 2025, Russian networks that used ChatGPT for political content. Bad Grammar generated Telegram comments about Russia, Ukraine and the Baltic States; Helgoland Bite produced German-language material before Germany’s 2025 federal election. OpenAI also said it shut accounts connected to Rybar in February.
Editorial analysis
This case is less about a high-reach burst than an effort to make later messages citeable. The sharper operational lesson is that the model use created a detection surface: OpenAI says scrutiny of AI-generated posts led it from content to the surrounding network. Account bans can disrupt a production layer, but the meaningful next test is whether the International Burke Institute site and linked channels persist or are rebuilt without the same generation trail. Their durability matters more than follower counts alone.
Loading discussion...
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.