How Can AI Help Advance Corruption Investigations?

It is cliche to say that artificial intelligence will impact nearly every industry, bringing changes both predictable and not. The question of how AI is affecting the anticorruption space has been covered by BriberyMatters, including these pieces on learning; these on the opportunities and risks of using AI in compliance programs; and several of my own touching on how AI intersects with procurement reform and the importance of data reliability. To continue this theme, I wanted to explore the role that AI can play in large-scale corruption investigations. Can AI be a reliable tool for investigators? How can AI be mobilized without sacrificing human judgement and expertise? What are the best applications of AI in such investigations?

To explore these questions, I spoke recently with Elisar Nurmagambet, an entrepreneur with deep AI expertise and professional experience tackling issues such as AML, financial crimes, KYC and counterparty risk for banks, sanctions evasion, shell company formation, databases and registries, the misuse of COVID-19 PPP loans, and more. One of the platforms his team designed was used by the UN to train international financial intelligence units to investigate corruption. Now, Nurmagambet runs Tesari, a company that aims to tap into the capabilities of AI to supercharge anticorruption investigations while still centering human judgment and expertise.

In Nurmagambet’s view, the appropriate role for AI is to be a “copilot, not an autopilot,” an approach known as “human in the loop.” He believes that AI should be a tool to augment, not replace, skilled and experienced human corruption investigators, journalists, expert witnesses, regulators and due diligence specialists. He explains that AI can significantly accelerate raw data collection, scraping thousands of databases, entity information, articles and other types of Open Source Intelligence (OSINT) in hours to find information about companies or individuals – work that would take a human weeks or months. This data creates a map or starting point that an investigator must then analyze and research more deeply. As Nurmagambet says, the investigator retains “control of every link.”

Nurmagambet also shared his perspective about why traditional corruption investigators need to embrace AI. As he put it, the current decade has given rise to the “most complex threat landscape in history,” with a polarized media landscape, disinformation and propaganda. Adversarial, corrupt and criminal actors, including sovereign states, he says, are taking advantage of those circumstances to augment their efforts to evade financial controls and sanctions. They are also weaponizing AI through deepfakes, impersonation attacks, social engineering and more. These AI-enabled threats can be undetectable for most users. Only the investigator trained to use those same tools can make better sense of the information, for example to judge whether a purported source is propaganda or disinformation.

Nurmagambet also shared his skepticism about one way in which AI is being marketed in the anticorruption space: what he terms “one-click comprehensive investigative reports” as a form of due diligence. The sales pitch is often made to corporate managers, based on cost savings, staff reductions and efficiency, rather than to legal departments. He feels such tools put clients at risk, particularly if they lack deep anticorruption investigative experience. Those risks, he says, “don’t outweigh the cost savings,” explaining that “decisions made based on faulty information do not stand up to evidentiary standards.” Comprehensive screening, he stresses, cannot be fully automated, as automated screens can generate false positives or miss pertinent information.

In sum, even as an AI expert himself, Nurmagambet has a deep respect for human instincts, judgement, integrity, decision making, and the power of the human analytical mind, particularly in the realm of countering corruption and financial crime. Even when AI reaches the level of “superintelligence,” he says, it cannot replace humans who “live and breathe” these issues, who understand concepts like chain of custody and evidentiary standards, and can interpret data and set aside bias to “follow the truth where it leads.” Given the wide range of threats posed by corruption in the global economy and political system, he says, “these investigators have the most important profession in the world today.”

Marc Schleifer

Governance, Democracy and Economic Development Expert