OpenAI says Russian-linked ChatGPT accounts were used to build a covert influence network designed to spread pro-Russia narratives while concealing their origins.

A cluster of ChatGPT accounts originating from Russia was used to support a covert online influence campaign aimed at spreading misinformation. OpenAI discovered the activity while investigating AI-generated social media posts, which led it to a “much broader influence operation,” according to a recent blog post.

OpenAI said the campaign centred around a website containing copied and misattributed academic work, as well as a “sovereignty” index that portrayed Russia in a positive light and was designed to disguise the operators’ Russian origins.

OpenAI has clarified that it does not provide access to its models in Russia. However, the company said the operators used VPNs to disguise their location while carrying out the activity.

Previously, OpenAI had shut down ChatGPT accounts connected to Rybar, a pro-Russian media company funded by Russian state corporation Rostec and reported to have worked with members of the Russian intelligence services.

The Aim Behind The Campaign

The operators wanted Russian agents to generate social media comments for posting across Telegram, X, Facebook and LinkedIn. They also instructed ChatGPT to remove linguistic clues that could reveal the content’s Russian origins.

The AI-generated content promoted an organisation called the International Burke Institute (IBI), which featured articles copied from genuine academic writing and published under false attribution.

techMore from Tech

OpenAI said the operation had limited immediate reach, but its significance lay in the infrastructure it had built. The AI-generated posts promoted what appeared to be a credible institution featuring purported experts, republished academic research and a proprietary risk index.

The company said the case highlights how influence actors can use AI to build authority, conceal the origins of favoured narratives and create assets that can be scaled over time. At the same time, OpenAI noted that the use of AI can also make such operations more vulnerable to detection.