Russia Ran a Covert Influence Campaign Through ChatGPT

OpenAI has shut down a Russia-linked influence operation that used ChatGPT for two jobs at once: writing pro-Kremlin content and scrubbing the linguistic fingerprints that would give its origin away. The company disclosed the takedown in a threat report published on August 25, announcing that it had banned the cluster of ChatGPT accounts behind the campaign.

The operation was built around a fake think tank called the International Burke Institute, which presented itself as based in Israel while actually being run from Russia through VPN connections. The intellectual output was thin: 34 of the institute's 36 articles, published between September 2025 and May 2026, were copied from other sources. Around that hollow core, the operators generated pro-Kremlin and anti-EU social media content in German and Russian, spreading it across X, LinkedIn, Facebook, Substack and Telegram. A German-language Telegram channel called "Lahme Ente" and roughly a dozen affiliated channels gathered 10,000 to 20,000 followers each. The target list spanned Germany, the United States, France, Poland and Turkey.

The model as a disguise, not just a pen

The account instructions OpenAI recovered are the most instructive part of the report. Beyond producing posts, operators asked the model to remove wording that might reveal a Russian author behind the text. Propaganda that reads natively used to require staff who genuinely knew the target language and culture; a chat window now substitutes for that entire desk. That is the real cost collapse this report documents.

Reach, measured honestly

OpenAI's own assessment of the campaign's impact is modest. Individual posts drew very few views, and the operation rates a category 3 on the Brookings six-point breakout scale, meaning it built scalable infrastructure without achieving broad reach. The honest reading is not reassuring so much as clarifying: the barrier to running a multi-platform, multi-language influence operation has dropped to nearly zero, and the surviving bottleneck is distribution, not production. Pew's finding that over a third of recent webpages carry an AI fingerprint describes the same flood from the other shore.

What to do with this if you run a business

Treat information hygiene as an operational practice rather than a civic virtue. Verify the source of any striking industry claim before it enters a deck or a decision, teach teams that a polished, native-sounding article is no longer evidence of a real institution behind it, and watch for coordinated-looking content waves around your own brand. Fake think tanks have become cheap to build. Trusting content without checking who built it is what has become expensive.

Sources: OpenAI, The Decoder