Eduardo Salcedo-Albarán, founder and director of the Vortex Foundation, spoke with City Journal associate editor Daniel Kennelly about his work and his latest book, Super Network of Corruption in Venezuela: Kleptocracy, Nepotism and Human Rights Violation.
How did you find a career in analyzing and combatting corruption and transnational criminal networks?
Since childhood, I have been interested in understanding crime and why someone is willing to hurt others, sometimes violently, to achieve a personal goal. I think that this curiosity is related to growing up in Bogotá during the 1980s, when Pablo Escobar was at war against Colombian “authorities.”
Every night, my family and I watched the news about Escobar murdering politicians, public servants, and journalists. Fortunately, we were safe during those days, but I remember at least two bombs exploding several blocks away from our home. As it turned out, one of the bombs had detonated in a small shopping center where we used to buy fried chicken on the weekend; the other had exploded where we shopped for groceries. The explosions happened during the Mother’s Day celebration in 1990.
I began asking myself why and how someone could do these things. And that’s what I ask myself every day, for example, when analyzing why and how a corrupt officer takes funds for himself from a public budget that was supposed to be used to feed the poorest. There is a short answer, related to uncontrollable and limitless ambition; but the longer answer implies complex social structures, cognitive dynamics, and institutional conditions—a list of interconnected variables that I hope to keep researching.
Could you describe how you apply artificial intelligence to this work?
I founded the Vortex Foundation in 2011, in Colombia, to develop innovative methods for analyzing complex social phenomena such as criminal networks. During our first project that year, we created algorithms that systematized and visualized qualitative data from an ongoing transitional justice process (2005–2013). We were able to describe massive networks of human victimization that happened in Colombia during the 1990s as the result of a war between left-wing guerrillas and right-wing narco-paramilitary groups.
Those algorithms evolved into a platform now called VORISOMA, which offered several tools for analyzing increasingly complex networks. In 2021, we began experimenting with machine learning; since 2022, we have published experimental results of what we call Machine Learning Models on Criminal Networks.
Over the past decade, we have analyzed and modeled criminal networks worldwide, resulting in a vast corpus of empirical data. Training Machine Learning Models with our data was a natural step. Today, these models enable us to extract semantic entities from massive text and to map Criminal Network Graphs in a fraction of the time and cost compared with conventional methods.
Large Language Models (LLMs) are expensive to train, so we still rely on third-party Natural Language Processing tools, for example, to locate thousands of individuals and companies involved in money laundering worldwide. However, we plan to train our LLMs that specialize in identifying criminal networks as soon as possible. Without these tools, understanding the complexity of global crime is impossible.
Your latest book looks at Venezuela’s “super network of corruption.” What makes Venezuela’s situation extreme?
Venezuela is empirically known as the worst corruption case worldwide. Before we modeled the super network of corruption in Venezuela, Brazil’s “Lava Jato” was the biggest known case of corruption. Eleven of the largest Brazilian companies formed a cartel to manipulate public tenders. This initially started in Brazil but later spread across Latin America. The cartel paid large bribes and financed candidates running for the presidency and local administrations in most Latin American countries. The model we generated and published about the Lava Jato corruption network involved 934 nodes/agents—individuals and companies—that established 2,782 documented interactions. The super network of corruption in Venezuela, by contrast, currently involves 10,300 nodes/agents, establishing 18,000 interactions across every continent. The amount of money stolen from the Venezuelan people is astonishing, and the network’s complexity surpasses the institutional capacities of any judiciary.
The transnational case of Venezuelan corruption is a good example of how technology and artificial intelligence are essential to understanding this phenomenon. Machine Learning Models and Large Language Models have been essential to understanding the money-laundering paths related to this case.
Venezuela was once the envy of the region, thanks to its having the largest oil reserves in the world, but during the past two decades the public budget has been looted. According to Transparency International’s 2022 Corruption Perceptions Index, Venezuela ranks among the last three countries, with a score comparable to war-torn nations like Syria and Somalia. Most public and institutional areas of the country have collapsed, from public health and transportation to infrastructure; 7.8 million Venezuelans have fled the country due to extreme poverty.
Can you tell us about your upcoming book?
During the past decade, we have insisted that technology is essential to understanding the scale and effects of complex criminal networks. Though sophisticated software exists for visualizing networks, most public entities investigating and prosecuting complex crimes, including corruption, do it through archaic and slow methodologies. Likewise, the quality of empirical data used to craft criminal-justice policy is often outdated and extremely partial, even naïve.
Judiciaries focus on the visible elements of criminal networks while omitting the underlying economic and political structures because they lack the concepts and real-time methodologies for understanding criminal systems. A huge gap exists between the technologies and methodologies used in the private versus the public sector. Meantime, criminal networks adopt new technologies and evolve at rates with which most bureaucracies struggle to keep pace. With the rise of AI, the gap will widen.
The book aims to highlight the importance of utilizing AI to comprehend complex systems such as criminal networks. Written with Luis Garay, my colleague of 15 years, the book will provide a comprehensive explanation of Machine Learning Models on Criminal Networks and emphasize the significance of empirical data in training such models. If policy designers, decision-makers, journalists, prosecutors, and judges, among others, do not understand and use the power of AI, the concentrated power of a few lawful and unlawful social players will only increase, generating dramatic distortions.
Top Photo: MR.Cole_Photographer/Moment/Getty Images