OpenAI says it has disrupted and banned a cluster of ChatGPT accounts that very likely originated in Russia and were used to support a covert influence operation promoting a self-described Israeli “expert community” called the International Burke Institute, or IBI.
The operation combined AI-assisted social media activity with a broader effort to manufacture credibility: a website containing copied and misattributed academic work, purported experts, multiple social media channels, and a so-called “sovereignty index” that portrayed Russia favorably while criticizing Western countries.
According to OpenAI, the campaign appears to have reached relatively small audiences. But its infrastructure tells a more important story about the evolving role of artificial intelligence in influence operations.
The lesson is clear: AI does not need to create an entire disinformation campaign on its own to make the campaign more effective. It can help operators write, translate, adapt, distribute and scale narratives while hiding where those narratives really come from.
From AI-Generated Posts to a Wider Influence Operation
OpenAI said its investigation began with AI-generated social media activity and eventually uncovered a broader operation centered on the International Burke Institute.
The operators prompted ChatGPT in Russian to generate social media comments, many of which were produced in English. According to OpenAI, they specifically asked the model to remove linguistic clues that could reveal their Russian origins.
The resulting content was distributed across several major platforms, including X, LinkedIn, Facebook, Substack and Telegram.
Because OpenAI does not permit access to its models from Russia, the company said the operators used VPNs to access the platform.
The IBI website itself was registered in February 2025 and claimed to represent an organization based in Israel. However, OpenAI said many articles published on the site were copied from genuine academic work, sometimes with false attribution. Other material appeared to have been written by a Slavic-language speaker and translated into other languages.
OpenAI stressed that the articles hosted on the website were not generated by its models. ChatGPT’s main role in the operation, according to the company’s investigation, was supporting the generation of promotional and social media content.
The “Sovereignty Index” and the Manufacturing of Authority
One of the campaign’s most distinctive features was an apparent attempt to create intellectual legitimacy around its messaging.
OpenAI identified the operation’s “sovereignty index” as a central component of the campaign. The index was promoted by the supposed think tank and presented as a way of evaluating countries, while casting Russia in a favorable light and criticizing Western nations.
That approach matters because modern influence operations increasingly go beyond simple propaganda posts.
A fake profile can be created in minutes. A credible-looking institution takes more work.
The IBI operation appears to have invested in the second approach: building a recognizable brand, publishing research-like content, associating itself with purported experts, republishing academic material and promoting an apparently proprietary analytical product.
This is where AI becomes particularly useful to influence operators. Artificial intelligence can help produce a constant stream of supporting content around a broader operation, making an institution or narrative ecosystem appear more active, responsive and internationally connected.
OpenAI described the operation’s significance as lying less in its immediate audience than in the infrastructure it had constructed.
Limited Reach, but an Important Warning
OpenAI assessed the campaign’s immediate impact as limited.
Typical social media posts received relatively low numbers of views, and the official IBI accounts had relatively small subscriber counts. Some Telegram channels associated with the broader activity, however, attracted larger audiences, generally between 10,000 and 20,000 subscribers.
Using the Brookings Breakout Scale, OpenAI assessed the operation at the lower end of Category Three, meaning it operated across multiple platforms and showed some signs of reaching authentic audiences.
But cybersecurity and influence operations are not always defined by their current scale.
Security teams regularly investigate infrastructure before it becomes a major incident. The same principle applies here.
A network of websites, social accounts, channels and apparently credible institutions can be built quietly and then expanded when a political crisis, conflict, election or geopolitical event creates an opportunity.
The operation disrupted by OpenAI demonstrates how influence infrastructure can be developed before its operators decide to scale it.
AI Is Becoming Part of the Influence Supply Chain
Perhaps the most important takeaway is that AI was not the entire operation.
It was one component.
OpenAI has repeatedly highlighted this pattern in its reporting on malicious uses of AI: threat actors tend to combine AI models with more traditional tools and infrastructure, including websites, social media accounts and other digital platforms.
That distinction is important.
The biggest risk may not always be an AI system independently creating a sophisticated propaganda network. The more immediate challenge is that AI can make traditional influence work cheaper, faster and easier to adapt across languages and audiences.
An operator can use AI to draft a post, rewrite it for a different audience, translate it, generate multiple variations and maintain activity across several platforms.
For defenders, this means that identifying AI-generated content alone is not enough.
The real question is:
What larger operation is that content supporting?
That is also where modern cybersecurity increasingly overlaps with threat intelligence, information integrity and digital risk monitoring.
A Familiar Pattern in Russia-Linked Influence Activity
The latest disruption is not the first Russia-origin influence activity publicly documented by OpenAI.
The company has previously reported disrupting Russia-linked operations that used AI tools to generate multilingual content, political messaging and material targeting audiences in Europe, Africa and elsewhere.
Previous cases included activity linked to the “Stop News” operation, which used AI-generated content in English, French and Russian and promoted websites posing as news outlets in Africa and the United Kingdom.
Other OpenAI reports have documented Russia-linked operations generating political content related to Ukraine, NATO and European politics.
The new IBI case is notable because of the apparent effort invested in constructing an institution rather than simply operating anonymous accounts.
That represents a broader evolution in the threat landscape: influence actors are increasingly trying to build ecosystems of credibility, not just individual pieces of content.
Why This Matters to Organizations and the Cybersecurity Industry
Influence operations are often discussed as political problems. But they can also create risks for businesses, governments and technology organizations.
A sophisticated information operation can target:
- Corporate reputations;
- Public confidence in institutions;
- Executives and senior leaders;
- Geopolitical narratives affecting markets;
- Elections and public policy;
- Cybersecurity companies and technology vendors;
- Public trust during a cyber incident or data breach.
Imagine a major cyberattack against a government or enterprise.
Alongside the technical intrusion, threat actors could use coordinated social media accounts, fake experts, fabricated research and AI-generated commentary to spread competing explanations of what happened.
The result could be confusion at exactly the moment when accurate information is most important.
This is why cyber resilience increasingly includes more than firewalls, endpoint protection and incident response.
Organizations also need to understand how information about them can be manipulated online.
Why This Matters for the Middle East and Africa
The implications are global, and the Middle East and Africa are no exception.
The MEA region is strategically important in global politics, energy, technology, finance and international trade. Governments and businesses are also accelerating their adoption of AI and digital platforms.
That creates opportunity but also a larger digital information environment that can be exploited.
Organizations in the region should not assume that an influence campaign must explicitly target their country to affect them. Narratives can move across borders and languages, particularly when AI makes content easier to adapt and redistribute.
The region’s multilingual environment is another important factor. AI-assisted translation and content generation can allow the same narrative to be rapidly adapted for English, Arabic, French and other languages.
For governments, enterprises and media organizations, improving cybersecurity awareness and digital resilience should therefore include an understanding of coordinated inauthentic behavior, impersonation and information manipulation.
For additional global cybersecurity reporting, analysis and threat coverage, readers can explore CyberCory.
10 Recommended Actions for Security and Communications Teams
1. Treat information manipulation as a security risk.
Cybersecurity, communications, legal and executive teams should have a shared process for handling coordinated false narratives and impersonation campaigns.
2. Monitor your organization’s digital footprint.
Track fake domains, impersonated executives, fraudulent social media accounts and unusual narratives involving your organization.
3. Verify before amplifying.
Do not share sensational content about cyber incidents, geopolitical events or companies before confirming its origin and credibility.
4. Look beyond the individual post.
A suspicious post may be part of a wider network. Investigate linked websites, social accounts, domains and recurring narratives.
5. Strengthen executive and brand protection.
Senior executives and public-facing organizations are frequent targets for impersonation and manipulated narratives.
6. Train employees to recognize manipulation techniques.
Security awareness programs should cover not only phishing and malware, but also fake experts, impersonation, coordinated amplification and deceptive content.
7. Establish trusted communication channels before a crisis.
Organizations should know in advance how they will communicate verified information during a cyber incident or misinformation campaign.
8. Preserve evidence of suspicious activity.
Screenshots, URLs, account information and timestamps can help investigators identify coordinated campaigns.
9. Coordinate with platforms and relevant authorities when necessary.
Where impersonation, fraud or coordinated manipulation creates a material risk, organizations should use appropriate reporting and escalation channels.
10. Build threat intelligence around narratives, not just malware.
Modern threat intelligence should monitor actors, infrastructure and campaigns across the broader digital ecosystem.
The Bigger Picture: Trust Itself Is Becoming a Target
The most important lesson from this case is not simply that AI can generate social media content.
That was already known.
The deeper concern is how AI can support attempts to manufacture trust.
A fake institution can look more active when AI helps generate regular content. A coordinated network can appear more international when AI helps translate messages. A fabricated expert community can seem more credible when it is surrounded by research-like material, commentary and social media engagement.
In other words, AI can act as an accelerator for influence infrastructure.
OpenAI’s disruption of the Russia-linked campaign also demonstrates the other side of the equation: AI companies themselves can use signals from model abuse to investigate and expose broader malicious activity.
According to OpenAI, the supporting use of ChatGPT in this case ultimately helped lead investigators toward the wider operation.
Conclusion
OpenAI’s disruption of the International Burke Institute-linked operation is a reminder that the next generation of influence campaigns may not always announce itself through viral propaganda.
Some will attempt to look legitimate.
They may build websites, publish research-like content, create analytical indexes, establish social media channels and present themselves as independent institutions.
AI can help make that process faster and easier but it does not remove the human infrastructure behind the operation.
For security professionals, governments, businesses and media organizations, the challenge is to look beyond individual pieces of suspicious content and examine the ecosystem supporting them.
The campaign identified by OpenAI may have had limited immediate reach, but its elaborate structure offers a valuable warning: in the age of AI, defending against influence operations increasingly means defending the systems of trust that shape how people decide what is real, credible and worth believing.
The bigger message for the global cybersecurity community is that AI-powered influence operations should not be viewed only as a problem for social media platforms. They are becoming part of the wider digital threat landscape and organizations will need to prepare accordingly.
