Scams scale faster than trust can recover.
Bogotá, March 2026
Latin America’s surge in digital fraud is not a passing wave of opportunistic crime. It is the predictable outcome of rapid digitisation colliding with uneven security maturity, expanding real-time payments, and a consumer base that moved online faster than institutions rebuilt their defensive muscle. Infobae’s reporting frames the problem in exactly those terms: more e-commerce, more mobile banking, more exposure, and fraud techniques that now evolve at platform speed. Regional assessments from bodies such as the Inter-American Development Bank and the Organization of American States have repeatedly warned that while cybersecurity capacity is improving, the region still faces structural gaps in preparedness and resilience. The result is a high-friction reality: the digital economy grows, and so does the surface area for manipulation.
What is changing in 2026 is not only the volume of scams, but their quality. Fraud is moving from crude phishing to psychologically engineered operations that exploit identity, urgency, and imitation. Generative AI has accelerated this shift by lowering the cost of persuasion. A scammer no longer needs perfect language skills to write convincing messages, or a large team to run campaigns at scale. With AI tools, they can localize content, personalize outreach, and vary scripts fast enough to outrun traditional rule-based defenses. UNESCO has warned that deepfakes and voice cloning are eroding trust by making impersonation cheap and believable with minimal source material. This is not just a “tech risk.” It is a trust crisis that targets the human layer of every system.
Latin America is especially exposed because of how much economic activity is concentrated in a few channels: messaging apps, mobile banking, marketplace platforms, and informal payment behaviors that blur personal and commercial communication. When fraud moves into the same channels people use to speak with family, coordinate work, and receive services, suspicion becomes a daily posture. The region’s fraud ecosystem thrives on this proximity. If a scam arrives through a channel that feels personal, the victim’s guard drops, and the criminal gains the crucial advantage of time. In modern fraud, time is the real currency. The scammer tries to compress decision-making into a few minutes before verification kicks in.
The most dangerous acceleration is impersonation. Reuters has documented cases in the region where criminals used deepfake content in advertising scams, including a Brazilian operation that leveraged AI-generated celebrity videos to push fraudulent offers and collect payments at scale. Even when the amounts per victim are small, the model works because it exploits “statistical immunity”: victims feel embarrassed, assume the loss is too minor to report, or do not know where to report. That underreporting is not a side issue. It is a structural enabler. It allows fraud to scale quietly while data quality stays poor, making it harder for institutions to measure the real size of the threat.
This is where AI enters the story as both weapon and shield. Banks, fintechs, and payment processors in Latin America are increasingly deploying machine learning systems to detect anomalies in real time, flag suspicious account creation, and identify mule networks that move stolen funds across accounts. The Dialogue has reported that financial services firms across Latin America and the Caribbean are adopting AI for fraud detection at scale, driven by the simple math that manual review cannot keep up with transaction velocity. AI can learn patterns that rule sets miss: unusual device fingerprints, behavioral inconsistencies, transaction sequences that resemble laundering, or network-level signals that indicate coordinated activity rather than isolated incidents.
But AI is not a miracle, and the region’s defensive success will depend on governance more than algorithms. The Financial Stability Board has warned that AI adoption in finance introduces vulnerabilities: third-party dependencies, cyber risks, model risk, and governance gaps that can create correlated failures. In plain terms, if many institutions buy similar models, train on similar data, and outsource similar components, they can become predictable to attackers and fragile to systemic shocks. Fraud prevention becomes a competition of adaptation. Attackers probe the edges of models, find blind spots, and pivot. Defenders retrain, refine, and tighten thresholds. This is why the right metaphor is an arms race, not a one-time upgrade.
There is also a political economy issue beneath the technology. In Latin America, fraud defenses are unevenly distributed. Large banks can afford sophisticated detection and dedicated teams; smaller institutions and merchants often cannot. This creates a displacement effect: criminals adapt by targeting the weakest nodes, smaller merchants, microbusinesses, under-protected consumers, and platforms where identity verification is thinner. As defenses improve in one place, fraud migrates to the next. That is why a purely institution-by-institution approach fails. Effective reduction requires shared intelligence, coordinated reporting, and faster takedown mechanisms across the ecosystem.
The human layer remains the hardest to defend, precisely because it is the most flexible. AI-driven scams increasingly use emotional manipulation rather than technical exploitation: fake customer support agents, “urgent” account warnings, voice messages that sound familiar, and staged narratives designed to bypass skepticism. The best technical defense still struggles when a victim is convinced to authorize the transfer themselves. This is the pivot that scammers rely on: turning security into consent. Once the victim approves the action, it becomes harder for banks to reverse and easier for criminals to argue that the transaction was “legitimate.” In this sense, education and friction are not annoyances. They are defensive design. A small pause, a second factor, or a verification step can break the scam’s timing.
What the region is learning, sometimes painfully, is that fraud prevention is becoming a core infrastructure problem. It cannot sit only in compliance departments or in customer service scripts. It has to be engineered into onboarding, authentication, payments, and dispute resolution. It also requires a cultural shift: reporting scams should be normalized and made easy, because silence is a multiplier for criminals. The more underreported fraud becomes, the more the ecosystem trains itself to accept it as background noise.
Latin America’s digital fraud boom, then, is not just about criminals getting smarter. It is about systems becoming more complex, more interconnected, and more dependent on trust. AI will help defend, but only if it is paired with strong governance, shared intelligence, and design choices that respect how real people behave under stress. The region is not only fighting fraud. It is fighting for the credibility of its digital economy.
Phoenix24: clarity in the grey zone. / Phoenix24: claridad en la zona gris.