Enhancing Financial Fraud Prevention Through AI-Driven Detection

Q: What was the business opportunity that led to the creation of Lynx and how has your position in the market evolved since then?
A: We were born out of an unsolved need in the financial sector. In the mid-1990s, while we were developing neural network models at university, a card processor in Spain asked IBM for a real-time fraud prevention tool. Lacking a proprietary solution, IBM turned to our team, which led us to develop a system capable of analyzing the risk of credit card transactions in a matter of milliseconds. This was the first step towards creating an AI platform applied to financial security.

What started as a non-profit university project evolved during the following 25 years by reinvesting profits in innovation and development. During this period, we collaborated with large financial institutions such as Grupo Santander, and Cielo (formerly Visanet Brazil), consolidating our solutions in multiple markets. In 2023, the need to scale and operate with greater autonomy led to the founding of Lynx as a commercial company, leveraging decades of experience in AI and fraud detection to offer advanced solutions to the financial sector.

Q: What distinguishing features set Lynx apart from other players in the market that also offer fraud prevention solutions?
A: First, our platform is highly flexible and configurable, allowing us to adapt the application to new customer requirements in a matter of hours, without the need for code modifications. This gives us a significant operational advantage over rigid solutions that require additional development. Second, our AI is constantly updated with real-time data, avoiding the deterioration of models trained with historical information. In an environment where payment methods and fraud patterns are constantly evolving, this daily self-learning capability allows us to maintain accurate and effective detection. Finally, we operate in a real response time of 50 milliseconds, ensuring seamless integration with any financial institution's authorization systems without impacting their performance.

Q: Lynx’s platform uses daily adaptive models. How does this approach improve fraud prevention and what advantages does it offer over traditional detection methods?
A: Unlike traditional models, which rely on historical data and require periodic retraining, our adaptive models are updated daily based on real-time feedback. This allows for more accurate detection of new fraud tactics without the delays associated with the obsolescence of previous models. By continuously adjusting parameters, we ensure effective protection against emerging threats, optimizing response without compromising processing speed, not only improving fraud prevention, but also reducing false positives and maximizing our clients' operational efficiency.

Q: Lynx prides itself on its ability to process up to 2,400 transactions per second, how is that efficiency achieved without compromising accuracy, speed, and user experience?
A: This is possible thanks to an architecture designed to scale with customer size without compromising accuracy. To achieve response times of less than 50 milliseconds, we developed our core in C, a high-performance language used in operating systems and critical applications. This allows us to optimize processing without losing reliability, ensuring that transactions are validated in real time with maximum accuracy.

Our platform achieves a fraud detection rate of 70%-80%, with a 0.2% impact on legitimate transactions. This balance allows financial institutions to minimize losses without significantly affecting the user experience. In addition, we adapt the level of detection according to the risk appetite of each client, ensuring that the measures implemented are effective without generating unnecessary friction.

Q: What technological challenges must the company overcome to maximize performance?
A: The challenge is maintaining performance in both cloud and on-premise environments, ensuring the same efficiency in both architectures. The cloud gives us virtually unlimited scalability, while in traditional data centers we optimize resources to maximize performance within physical constraints. Our engineering teams develop and maintain both versions of the platform with a unified approach, ensuring consistency and stability across any infrastructure.

Q: How do you ensure that Lynx’s mule account detection capabilities are accurate and do not generate false positives while protecting financial institutions from complex fraud?
A: We perform an analysis of financial behavior, rather than identifying the origin of the account. Our platform monitors incoming and outgoing transaction patterns to identify irregular movements and generate real-time alerts. While false positives are inevitable, we work on calibrating risk models and thresholds to keep them at optimal levels, minimizing disruptions for legitimate customers without compromising the security of the institution.

Q: What will be the most disruptive trends in the use of AI for financial security in the coming years?
A: There is no single AI solution that solves all problems. Each technology has its own approach and applicability. The future of AI in financial security lies in combining different models to achieve greater accuracy. This collaborative approach, similar to consulting multiple opinions to make a better decision, will enable more effective fraud detection and reduce false positives. The key will be to integrate technologies such as supervised and unsupervised ML, neural networks, and real-time behavioral analysis to anticipate threats before they materialize.

Q: How can companies collaborate with authorities to generate policies and tools to fight financial fraud?
A: Regulators play a key role in protecting end users, who are the most vulnerable part of the financial system. Therefore, they should encourage financial institutions to invest in security technology, since the responsibility for mitigating fraud falls on them. In many cases, establishing regulations that assign responsibility for fraud to the institutions, when it occurs due to failures in their procedures, generates an incentive to improve security. Collaboration between companies and authorities should focus on sharing information on new threats and developing standards to ensure a coordinated response to financial attacks.

Q: What are your plans to further strengthen your presence in Mexico and Latin America in 2025?
A: We worked with clients in Latin America without local teams for 25 years. Now, we see a great opportunity in the market and are investing in local talent to strengthen our presence in the region. In Mexico, we already have staff and plan to expand our network of relationships with financial institutions to tailor our solutions to the specific needs of the market. Investing in local teams will enable us to better understand the dynamics of the sector and accelerate the adoption of our technologies in the region.

Q: What steps is Lynx taking to expand its range of solutions and adapt to the changing needs of the market?
A: The basis for success lies in listening to customers and staying at the technological forefront. There is no point in developing technology without knowing the real challenges of customers. Lynx combines academic research with practical experience to apply the right solutions to each problem. We collaborate with universities to keep abreast of advances in AI, while maintaining an ongoing dialogue with our customers to tailor our solutions to their specific needs.