Malik Haidar is a seasoned veteran in the cybersecurity trenches, known for his sharp analytical eye regarding how multinational interests intersect with digital warfare. With a career built on dismantling complex threats for global corporations, Haidar has become a leading voice in understanding the bridge between technical intelligence and business strategy. Today, we delve into a recent discovery where OpenAI dismantled a sophisticated Russian operation that utilized generative AI to breathe life into a faux think tank. Our conversation explores the tactical shift from crude bot networks to the creation of “manufactured authority” and the technical cat-and-mouse game of regional access and linguistic masking.
How do influence operations effectively use VPNs to bypass regional access restrictions, and what specific patterns do security teams identify when these actors try to mask their origins?
In the case of this Russian cluster, the actors utilized virtual private networks to circumvent geographic blocks that would typically prevent them from accessing tools like ChatGPT. By routing their traffic through servers in permitted regions, they attempted to appear as legitimate users, but their digital footprints eventually gave them away. Security analysts look for more than just IP addresses; they examine the behavioral nuances, such as when a user instructs the AI to specifically conceal linguistic clues that might reveal a Slavic or Russian provenance. In this operation, the actors were meticulously directing the model to generate content that appeared natively Western, yet the underlying metadata and the systemic nature of the account creation pointed back to a coordinated effort. This constant tension between technical obfuscation and the behavioral patterns of the operators is where we usually find the first cracks in their armor.
The International Burke Institute presents a fascinating case of “manufactured authority”—could you explain how this entity was constructed to look credible to an unsuspecting audience?
The International Burke Institute, or IBI, was essentially a digital ghost designed to look like a prestigious “expert community” based in Israel. Registered as recently as February 2025, the website served as a polished front, claiming to host a global network of specialists in economics and international relations. To build this facade of legitimacy, the operators didn’t just use AI-generated text; they actually copied and misattributed real academic work to fill their pages. This approach makes the site feel grounded in reality, as a casual reader might see familiar citations or high-level academic jargon and assume the institution is a long-standing authority. It is a sophisticated form of social engineering where the goal isn’t just to spread a single lie, but to build an entire environment where those lies can feel like expert-vetted truths.
A central piece of this campaign was the “Sovereignty Index.” What are the technical and psychological mechanics behind using such a complex metric to influence public opinion?
The Sovereignty Index, also marketed as the Burke Index, was a clever psychological tool that used the veneer of data science to project a specific geopolitical narrative. By claiming to measure seven distinct indicators—including political, economic, technological, information, cultural, cognitive, and military sovereignty—the operators created a framework that felt objective. Each area was allegedly calculated using “equalization coefficients” on a 100-point scale, leading to a cumulative score between 100 and 700. When they released their findings, they placed the U.S. at 650.9 and China at 649.1, but crucially positioned Russia at 601.4, describing it as a world power with “full digital and military nuclear sovereignty.” This use of specific, seemingly scientific numbers is intended to bypass a reader’s natural skepticism by overwhelming them with a structured, data-driven argument that favors the operator’s home country.
While the overall audience reach was described as relatively small, the infrastructure was surprisingly robust. How did the operators use AI tools to manage and scale their digital assets, such as the “American Observer” channel?
The infrastructure allowed a small group of actors to act like a much larger media organization by using ChatGPT as a force multiplier for content production. For instance, one user utilized the tool to generate a professional-looking profile picture for a Telegram channel dubbed the “American Observer” and frequently requested Russian-language summaries of the channel’s activity to keep the controllers informed. On platforms like Telegram, they managed to amass between 10,000 and 20,000 followers per channel, which is a significant number for an operation that was essentially built on a foundation of fake academic work and machine-translated reports. Even though the AI was only used to produce isolated promotional posts on platforms like X, Facebook, and LinkedIn, those posts were the “hooks” that led people back to their more elaborate, credible-appearing website. This shows that the true danger isn’t just in the volume of AI content, but in how it is used to funnel users into a broader, more deceptive ecosystem.
What is your forecast for the evolution of these “think tank” style influence operations?
We are moving into an era where influence operations will prioritize quality and perceived authority over the sheer volume of spam we saw in previous years. In the near future, I expect these actors to become even more adept at using AI to build “persistent assets”—digital entities that have years of back-dated history, interconnected social profiles, and deep libraries of synthesized research. As we have seen with the IBI, the significance lies in the infrastructure; they are building the “plumbing” of disinformation that can be turned on or off at will to support specific narratives. However, the irony is that the very same AI tools they use to scale their deception also leave behind distinct signatures that allow companies like OpenAI to detect and neutralize them. The battleground will increasingly be defined by who can better hide or identify these “synthetic authorities” before they can take root in the public consciousness.
