OpenAI has banned a cluster of ChatGPT accounts that it says originated in Russia and were used to run a covert online influence campaign built around a purported think tank.[2][1][5] The company disclosed the action in a Tuesday, Aug. 25 blog post, noting that the accounts were being used to promote the International Burke Institute (IBI), a self-described “expert community” based in Israel.OpenAI[2][1][5] According to OpenAI, the operation relied on its AI models to generate social media posts and comments that operators then published across multiple platforms under the guise of legitimate analysis.[2][1][13]
OpenAI said the network accessed ChatGPT via virtual private networks to bypass access restrictions, prompting the model in Russian while asking it to produce English and German comments tailored to different audiences.[2][13] Those comments were then posted on Substack, Telegram, X, Facebook and LinkedIn accounts that appeared unaffiliated with each other, creating the impression of organic engagement around IBI’s work.[2][13][1] The core of the campaign revolved around an IBI-branded website hosting copied and misattributed academic work and a “sovereignty index” that systematically rated Russia more favorably than Western countries while supplying talking points for social content.[2][1][14]
OpenAI assessed that the effort was designed to launder pro-Kremlin narratives through a veneer of academic authority, with IBI presenting itself as an Israeli expert community while obscuring the Russian origin of its operators.[2][5][13] In its report, the company noted attempts to hide linguistic cues of the authors’ native language and to reuse the same AI-generated text across multiple personas, a hallmark of coordinated inauthentic behavior.[2][13] Researchers who reviewed the content said many posts pushed themes aligned with Russian state media on issues from the war in Ukraine to European politics, but avoided explicit references to official propaganda outlets.CNBC[1][13]
OpenAI framed the takedown as part of a broader campaign to disrupt malicious uses of its models in information operations, following earlier actions against the Russia-linked “Doppelganger” network that used AI to generate anti-Ukraine social media comments and website copy in multiple languages.OpenAI[10][7] The company has also detailed Russia-origin efforts it dubbed “Stop News,” which relied on generative models to produce multilingual articles and video scripts targeting audiences in Ukraine, Africa and the West.[4][6][9] Separate OpenAI reporting describes bans on ChatGPT accounts tied to Russian-speaking criminal groups experimenting with the model for malware tooling, underscoring the overlap between influence operations and more traditional cybercrime ecosystems.OpenAI[8][3]
In the IBI case, OpenAI emphasized that its models did not provide the operators with novel capabilities beyond what is already possible using publicly available tools, but warned that the scale, speed and linguistic flexibility of AI-generated messaging can amplify the reach of such campaigns.[2][1] The firm said it disabled all identified accounts, blocked associated payment instruments and shared relevant technical indicators with social media platforms and other industry partners to help them detect and remove related content.[2][8][1] OpenAI’s disclosure comes amid intensifying regulatory and public scrutiny of how frontier AI systems can be weaponized in elections and geopolitical conflicts, even when no conventional software vulnerabilities or malware are involved.[7][1]
Defenders tracking information operations will need to treat AI-assisted campaigns such as IBI as part of the broader threat landscape, watching for opaque “expert communities” that appear suddenly, publish highly polished but thinly sourced analysis and seed identical narratives across many platforms. In practical terms, that means closer collaboration between platform trust and safety teams, threat intelligence researchers and AI providers, along with more rigorous verification of think tanks and research outlets that seek to shape policy debates online.
References
- OpenAI bans Russian ChatGPT accounts used in covert …
- Disrupting a new covert influence campaign from Russia
- OpenAI bans ChatGPT accounts linked to Russian, …
- Operation “Stop News”: Russia-origin influence activity | OpenAI
- OpenAI shuts down Russia-linked ChatGPT accounts over ‘ …
- Operation “Stop News”: Recidivist influence activity
- In a first, OpenAI removes influence operations tied to Russia, China and Israel
- Cyber Operation: Russian-speaking malware tooling
- OpenAI bans some Chinese, Russian accounts using AI for evil
- Doppelganger – Russian influence activity targeting Ukraine – OpenAI
- Russia’s bot network used ChatGPT in a massive covert operation
- OpenAI bans Russian ChatGPT accounts used in covert …
