OpenAI has banned a cluster of Russian-linked ChatGPT accounts that were allegedly used to support a covert influence operation targeting audiences across multiple social media platforms.
The company said the accounts used ChatGPT to generate social media posts and comments that were distributed across Substack, Telegram, X, Facebook and LinkedIn.
OpenAI said the operators used VPNs to bypass access restrictions and attempted to conceal their Russian origins while promoting an organization called the International Burke Institute (IBI).
The investigation began with AI-generated social media content but eventually uncovered a broader operation involving a website, purported academic material and a sovereignty ranking designed to present Russia in a favorable light.
Russian Operators Used ChatGPT to Generate Influence Content
According to OpenAI, the banned accounts were used to promote IBI, which describes itself as an expert community based in Israel focused on economics, politics, sociology and international relations.
The company said the operators instructed ChatGPT to generate content in English while deliberately removing linguistic characteristics that could reveal their Russian background.
The campaign’s overall reach was relatively limited, although some of its Telegram channels reportedly attracted 10,000 to 20,000 followers each.
OpenAI said most of the AI-generated material was written in English and distributed through a network of social media accounts.
The activity was not limited to text generation. One ChatGPT account was also used to create a profile image for a Telegram channel called “American Observer.” The same account repeatedly requested Russian-language summaries of the channel’s activity.
International Burke Institute Website Raised Questions
The IBI brand is connected to a website registered in February 2025.
The site presents the International Burke Institute as a global community of experts covering subjects such as economics, politics, sociology and international relations.
OpenAI said it found no evidence that the site’s articles were themselves generated using ChatGPT.
However, researchers found that some articles appeared to have been copied from academic publications, in certain cases with incorrect or false attribution.
Other content appeared to have been written by a Slavic-language speaker and subsequently translated into English using machine translation.
OpenAI said the influence operation appeared to be distinct from previously documented Russia-linked influence campaigns.
One of its most notable features was the creation and promotion of a Sovereignty Index, also referred to as the Burke Index.
Sovereignty Index Promoted Russia Favorably
The IBI website describes its Sovereignty Index as a measurement based on seven areas:
- Political sovereignty
- Economic sovereignty
- Technological sovereignty
- Information sovereignty
- Cultural sovereignty
- Cognitive sovereignty
- Military sovereignty
According to the website, each category is assigned a score based on official national and international data sources, as well as expert assessments.
The seven scores are then combined to produce a cumulative score ranging from 100 to 700.
The index ranks countries according to their claimed level of national sovereignty.
At the time described by OpenAI, the United States ranked first with a score of 650.9, followed by China at 649.1, Switzerland at 610.7 and Russia at 601.4.
The report on Russia presents the country as having a high level of sovereignty and significant military and economic self-sufficiency while identifying demographic aging, technological innovation constraints and selective global integration as ongoing challenges.
It also portrays Russia as one of three global powers with full digital and military nuclear sovereignty.
OpenAI said the index was an important component of the broader influence effort because it provided a seemingly independent analytical framework that could be used to promote favorable narratives about Russia and criticize Western countries.
Operation Attempted to Build an Appearance of Authority
OpenAI said the significance of the operation was not primarily the size of its audience.
Instead, researchers highlighted the infrastructure established around the campaign.
The operation reportedly combined AI-generated promotional material with a website designed to appear credible, purported experts, republished academic research and a proprietary-looking international ranking.
This approach could help influence operators establish an apparently legitimate source that can later be used to distribute narratives at greater scale.
OpenAI said the operators used ChatGPT only for selected parts of the operation rather than relying on the AI system to produce the entire campaign.
That distinction is important because the broader infrastructure — including websites, social media accounts and purported research — existed outside the AI platform.
AI Used as a Supporting Tool
OpenAI characterized the operation as an example of how influence actors can incorporate generative AI into established influence techniques.
Rather than creating an entirely AI-generated propaganda network, the operators used ChatGPT as a supporting tool for producing promotional posts, comments, summaries and other content.
The company said this allowed the operators to accelerate parts of their activity while attempting to disguise the origin of the campaign.
OpenAI also noted an apparent contradiction: the use of AI helped the operators produce content at scale, but their interaction with the AI system ultimately provided investigators with information that helped expose the wider operation.
Campaign Had Limited Reach but Broader Strategic Goals
Although the campaign’s direct audience was relatively small, OpenAI believes its infrastructure could have provided a foundation for future expansion.
The combination of a branded institution, purported experts, academic-looking material and a proprietary index created an ecosystem capable of giving selected narratives an appearance of legitimacy.
This is particularly significant in influence operations, where the objective is not necessarily to reach the largest possible audience immediately.
Instead, operators may attempt to establish seemingly credible sources that can be repeatedly referenced and amplified across different platforms over time.
OpenAI Continues to Target Covert Influence Operations
The latest action demonstrates the growing role of AI platforms in investigations into coordinated influence campaigns.
OpenAI said it banned the accounts involved in the activity after identifying their use in the operation.
The company emphasized that the campaign’s use of ChatGPT represented only one part of a much broader effort.
The case also highlights an emerging challenge for cybersecurity and information-integrity researchers: identifying operations that combine AI-generated content with traditional websites, social media networks, fabricated authority and misleading research.
As generative AI tools become more widely available, influence actors can use them to streamline individual tasks without necessarily making AI the central component of an operation.
OpenAI’s investigation suggests that this supporting role can still be significant — and that traces left through AI services may also provide investigators with valuable clues about otherwise concealed influence networks.
