OpenAI has taken decisive action against a network of ChatGPT accounts connected to a covert influence campaign originating from Russia. These accounts were actively involved in crafting social media posts and responses aimed at promoting the International Burke Institute (IBI), an organization claiming to be a group of Israeli experts.
This campaign did not rely on malicious software or exploits but instead utilized generative AI to distribute content across platforms like X, LinkedIn, Facebook, Substack, and Telegram. The operators remained largely anonymous.
Uncovering the Influence Operation
OpenAI analysts were able to trace the activity back to a network characterized by AI-generated posts, copied academic articles, and false authorship. The network’s so-called sovereignty index praised Russia while criticizing Western countries that support Ukraine.
According to a report shared with Cyber Security News, OpenAI noted that although the audience of this campaign was limited, the infrastructure was designed for potential growth. This highlights how deceptive operations can cleverly blend real published work with misleading branding to mask their true origins.
Details of the Account Suspension
The accounts in question utilized ChatGPT in Russian but requested content in English, avoiding any linguistic clues that could reveal Russian origins. To bypass OpenAI’s restriction on Russian access, operators used virtual private networks.
After identifying the connection between these generated posts and the larger operation, OpenAI banned the cluster of accounts. These accounts were promoting IBI articles through named profiles and fake personas.
German-language content created by one operator criticized Ukraine, the EU, and Germany while advocating for closer ties with Russia. Another operator designed logos for Telegram channels targeting various countries, aligning with patterns seen in a broader Russian fake-news network.
Assessing the Campaign’s Impact
The IBI website, registered in February 2025, portrayed itself as an Israeli expert community. However, OpenAI found that most articles linked to it were plagiarized from other sources, misattributed to incorrect authors.
These articles framed Western nations as weakened while casting Russia in a positive light. Such fluent posts on familiar platforms can obscure the deception from unsuspecting audiences.
Readers are urged to approach accounts that frequently link to a single outlet or offer vague institutional claims with caution. Validating the source and authorship of quoted experts is essential in preventing the spread of misinformation.
OpenAI’s assessment indicates that while the campaign did not achieve widespread reach, it created a foundation that could be expanded. Verifying claims with primary sources and reporting misleading accounts are crucial steps to curb such influence campaigns.
In conclusion, fostering safe AI tool usage and recognizing low-profile influence activities are critical in today’s digital landscape. OpenAI’s actions serve as a reminder of the ongoing need for vigilance in combating misinformation.
