OpenAI says it has disrupted a covert Russian influence campaign that used ChatGPT to create social media content aimed at Western audiences. The company banned a cluster of ChatGPT accounts tied to the activity after identifying how the operators used the platform.
The campaign centered on pro-Kremlin narratives and the promotion of the "International Burke Institute" (IBI), described as a think tank supposedly based in Israel. Its reach was limited in some places, but OpenAI said the operation had built infrastructure that could have supported broader distribution over time.
How ChatGPT was used in the campaign
According to the source article, the operators accessed ChatGPT from Russia while using VPNs. They also asked ChatGPT to remove linguistic signs that could point back to a Russian origin.
That detail matters because the campaign was not only using AI to draft text. It was also using the tool to make the content appear less connected to the people creating it. In an influence campaign, that kind of concealment can make social media posts, institutional pages, and summaries look more native to the audiences they are meant to reach.
The main public-facing asset was the "International Burke Institute" (IBI). The IBI promoted a "sovereignty index" that presented Russia favorably while making Western countries appear weaker by comparison.
OpenAI also found problems with the institute's article output. According to OpenAI, 34 of 36 expert-linked IBI articles published between September 2025 and May 2026 were copied from other sources. Some used fake author credits.
Two examples show how the attribution issue worked. One article from Cambridge University Press was falsely credited to a professor at the University of Nottingham. Another piece from the Migration Policy Institute was credited to an Australian food chemistry professor.
Where the material appeared
The campaign pushed material across X, LinkedIn, Facebook, Substack, and Telegram. That mix gave the operators access to public social feeds, professional networks, newsletter-style publishing, and channel-based messaging.
Some of the content appeared in German on a Telegram channel called "Lahme Ente" ("Lame Duck"). That channel criticized Ukraine, the EU, and the German federal government, while also arguing for closer ties with Russia.
The operation also included a second operator who created logos for about a dozen Telegram channels. Those channels focused on Germany, the US, France, Poland, and Turkey. The same operator regularly asked for Russian-language summaries of channel activity.
Taken together, these details show a campaign built around repeated publishing rather than a single viral post. The structure included branded accounts, Telegram channels, copied institutional-style articles, and AI-assisted messaging across multiple platforms.
Small reach, but organized infrastructure
OpenAI said individual posts received very few views, and the official IBI accounts had low subscriber counts. That means the campaign did not appear to generate large direct audiences through those accounts.
The Telegram side looked different. Linked Telegram channels each reached 10,000 to 20,000 followers, according to the source article. That made Telegram a more significant part of the campaign's distribution network than the low engagement on individual posts might suggest.
OpenAI rated the operation at category three out of six on the Brookings Breakout Scale. In the source article, that rating means the campaign spread across multiple platforms and showed early signs of reaching real users.
The key point is not that this specific campaign had already become large everywhere. It is that OpenAI saw an elaborate setup that could have grown. A low-reach campaign can still matter if it has the accounts, channels, branding, and content workflow needed to expand later.
Why this case matters for AI and influence operations
This was not the first Russian influence operation OpenAI says it has disrupted. In June 2024, OpenAI exposed a Russian network called "Bad Grammar" that used its models to generate political comments on Telegram about Russia, Ukraine, and the Baltic States.
A year later, OpenAI identified "Operation Helgoland Bite," another Russian campaign. That operation used ChatGPT to produce German-language content ahead of Germany's 2025 federal election, attacking the US and NATO while promoting the AfD party.
The latest case adds another pattern: AI was used alongside copied articles, fake attributions, platform-specific posts, and Telegram channels aimed at several countries. The campaign did not depend on one format. It combined institutional presentation, social media distribution, and audience monitoring.
The source article also notes an important limit. These are only the cases OpenAI caught on its own platform. It adds that Chinese AI providers and open-weight models are freely available, making the actual scale of AI-powered influence operations likely much larger.
For readers, the practical takeaway is straightforward. AI tools can reduce the effort needed to generate posts, adapt language, summarize activity, and maintain many channels. That does not guarantee a campaign will gain traction, but it can make experimentation and scaling easier for operators willing to hide their origin.
OpenAI's action shows one response: identify abusive account clusters and ban them. The broader challenge is that influence campaigns can move between platforms, accounts, and AI systems. This case shows how quickly generative AI can become part of the operational machinery behind political messaging campaigns.