As of April 2023, there were 1,000 active fintechs in Latin America (LatAm) with a vast majority focusing on financial inclusion, tackling the issue of 70 per cent of the population not having access to formal financial services. Em Conversa looks to uncover what the future of fintech could look like in the region, following a $2.1billion valuation in 2022.
Identifying fraud in the age of artificial intelligence (AI) is much harder than it ever has been before. Fraudsters are more cunning and deceiving, taking advantage of those who either do not know or refuse to acknowledge the power of AI. However, while the tech can be very dangerous in the wrong hands, AI is also the best way to identify and combat this type of fraud.
After Southeast Asia, Latin America has the highest percentage of overall revenue lost to fraud (20 per cent), and one in five e-commerce transactions are declined as fraudulent – twice as many as anywhere else on the planet. Organisations need help to combat fraud or they risk losing a lot more. One firm doing this is Lynx Tech.
To find out more we sat down with Carlos Santa Cruz, chief technology officer, Lynx Tech.
Can you tell me more about the company and your role within it?
Twenty-five years ago, a few of us came together to create a new technology with a simple but important goal: to fight financial fraud using advanced AI and machine learning. We spent the next two decades founding and building Lynx Tech, working with tier-1 banks to co-develop our solution and get to where we are today. Now we are protecting over 300 million customers and processing billions of transactions per year across nine countries.
Since securing a Series A funding round in November 2023, Lynx Tech introduced its AI-driven software solutions to solve clients’ most significant fraud and financial crime challenges. Specialising in detecting and predicting financial behavioural patterns, we continue to develop solutions based on our self-learning AI technology.
Our daily adaptative model not only learns from what has happened in the past from a behaviour and transactional perspective but also analyses each day’s transactions to understand and prevent evolving fraudulent activity.
As attacks become more complex, so do the prevention measures, enabling us to monitor 2,400 transactions per second and identify and block fraud. We set the standard for accuracy, speed, and scalability for multinational financial institutions and payment providers around the mobile.
What are some AI fraud trends we’re seeing in Latin America?
Over the last year, there has been a 33 per cent increase in social engineering attacks across the continent, with notable spikes in Mexico, Brazil, Chile, and Argentina. Fraudsters are leveraging artificial intelligence (AI) tools like voice cloning and deepfakes to create highly convincing fake audio or video messages to deceive victims into transferring money or revealing sensitive information.
Latin America also saw a staggering 617 per cent increase in phishing attacks in 2023, with cyber criminals using AI to craft more realistic and targeted phishing emails. Brazil and Mexico were the most affected, recording millions of phishing attempts aimed at stealing financial data and credentials.
It’s important to remember that these attacks — stemming from a new and emerging technology — are taking place in an area struggling to fight back against other types of fraud. After South-East Asia, Latin America has the highest percentage of overall revenue lost to any kind of fraud (20 per cent), and one in five e-commerce transactions is declined as fraudulent — twice as many as anywhere else on the planet.
How do the Latin American AI and anti-fraud sectors compare to that of the rest of the world?
While fraudsters attempt to use AI for nefarious purposes, Latin America is firmly embracing the technology to fight back with advanced fraud detection and prevention tools that learn and adapt over time.
We’ve been working with financial institutions across the continent to build machine-learning models that analyse large volumes of transaction data and identify patterns indicative of fraudulent behaviour.
We work closely with our customers to ensure the daily improvement of machine learning models, via our proprietary daily adaptive model. This not only prevents data drift but significantly reduces the amount of fraud compared to traditional static models while minimising false positives. The old ways of fraud detection relied on static rules that could not keep pace with rapidly evolving fraud tactics.
This manual approach not only introduced significant operational risks by handling sensitive personal data but also failed to provide the agility and scalability demanded by today’s digital landscape.
In contrast, our AI-powered models continuously ingest new data streams to dynamically update fraud detection capabilities. This enables proactive responses to emerging threats while enhancing accuracy through constant retraining and optimisation. The automated process minimises the human error and data privacy risks inherent in legacy methods. As such, harnessing AI is pivotal to stay ahead of tech-savvy fraudsters exploiting the latest tools.
Behavioural biometrics
Other key AI-powered fraud prevention approaches gaining traction in the region include behavioural biometrics to detect account takeovers. This includes analysing factors like mouse movements, keystroke dynamics, and mobile sensor data to detect abnormal user behaviour. AI is also enhancing identity verification capabilities by cross-referencing biometric and behavioural data against known fraud patterns to detect synthetic identities used for account opening fraud.
With financial institutions embracing machine learning models and a range of AI-powered fraud detection techniques, Latin America is becoming increasingly well-positioned to get ahead – and stay ahead – of evolving threats.
What are some unique challenges associated with the region in the AI fraud space?
Latin America faces several distinct challenges in the AI fraud landscape, especially when compared to other regions. A major hurdle is the lack of robust cybersecurity strategies, with only 48 per cent of financial institutions and/or fintechs having a dedicated plan. This leaves many vulnerable to sophisticated AI-powered attacks like deepfakes and voice cloning.
The rapid digital adoption sweeping across Latin America is also presenting a challenge. Take e-commerce; according to Statista, e-commerce users in Latin America is forecast to reach approximately 185.5 million by the end of 2024. Come 2029, this figure could grow by 52 per cent, resulting in over 281 million users across the region.
As such, this has expanded the potential attack surface for fraudsters exploiting AI tools. Infrastructural gaps, talent shortages, and evolving regulatory frameworks have also hindered AI readiness in combating fraud.
Moreover, the diverse risk profiles and variations across Latin American markets necessitate highly localised, tailored AI fraud prevention strategies rather than one-size-fits-all solutions. Overcoming these challenges requires enhancing cybersecurity preparedness, cultivating AI talent, robust data governance, and public-private collaboration to outpace AI-savvy fraudsters operating in the region. It’s a big challenge, but one I’m confident we can rise to.
What is Lynx Tech doing to improve the AI security sector in Latin America?
Lynx Tech is at the forefront of pioneering AI capabilities to combat fraud across Latin America. Our Daily Adaptive Models represent a paradigm shift in fraud detection and prevention powered by machine learning.
Our advanced AI models are trained on massive volumes of data from our extensive partner network of leading financial institutions in the region. By continuously ingesting new transaction streams, the models dynamically adapt and learn to detect even the slightest deviations indicative of fraudulent behaviour patterns. A key advantage is their ability to accurately identify genuine fraud while maintaining exceptionally low false positive rates, minimising customer friction.
Our daily adaptive models leverage fully automated machine learning pipelines, reducing human error risks and enhancing data privacy compared to traditional manual processes. Robust controls – including intentional noise injection during training – ensure the models remain resilient and tamper-proof even under extreme conditions. Stringent testing has validated their capability to function effectively with only a fraction of features present.
Through diligent monitoring, we promote upgraded models only when their performance surpasses existing versions across key metrics like fraud detection rates and false positives. This capability to iterate rapidly with enhanced accuracy daily is pivotal for proactively combating fast-evolving fraud tactics.
By combining this continuous model optimisation with a localised strategic approach tailored to unique Latin American market nuances, we’re able to provide businesses with best-in-class, fully compliant, and future-proof AI fraud prevention. As digital transactional volumes almost exponentially increase, our solutions empower the region to harness AI’s potential safely and effectively, shielding against emerging threats.
Plans for the future (roadmap and growth plan)
Our vision at Lynx is to design a comprehensive platform that unites the fight against fraud, AML challenges, and cybersecurity threats under one roof. Backed by a formidable team, we’re poised to redefine and revolutionise the approach to secure financial transactions on a global stage.