AI-driven phishing attacks surge in Latin America
Phishing attacks powered by artificial intelligence surged across Latin America, with Kaspersky detecting 1.291 billion blocks in the 12 months prior to September 2025, an 85% regional increase. Criminals are exploiting Robotic Process Automation to operate phishing farms that send mass SMS messages and using deepfake technology to impersonate celebrities and authorities in social media scams.
What happened
Phishing attacks driven by artificial intelligence surged across Latin America in the 12 months prior to September 2025, with 1.291 billion phishing blocks detected by Kaspersky in the region, an 85% increase from the prior period (2023/2024).
Criminals are using Robotic Process Automation (RPA) technology to operate 'phishing farms'—multiple connected mobile devices controlled to send SMS smishing messages at scale. These tools, originally designed for corporate process optimization, are being exploited to deliver mass personalized fraudulent messages allowing attackers to reach thousands of victims within minutes with high sophistication and difficulty to trace.
Attackers are also using AI-manipulated audio and video (deepfake and deepvoice) featuring the images and voices of celebrities, authorities, and acquaintances primarily on social media to distribute scams, inducing victims to believe in false financial gain opportunities.
Reported impact
In Costa Rica, 7.9 million detections were recorded over 12 months, representing a 132% increase from the prior period and averaging 21,000 detections daily.
In the Dominican Republic, 4 million detections were recorded over 12 months, representing a 14% increase from the prior period and averaging 11,000 detections daily. Across Latin America, attacks averaged 3.5 million daily with over 2,400 per minute.
Index analysis
Kaspersky specialists identified AI as the principal driver accelerating phishing attacks, used to refine and automate massive attacks at scale. The exploitation of RPA technology allows scammers to minimize cost and effort while distributing millions of messages.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.