Mexico City — Identity fraud powered by artificial intelligence now accounts for 23.3% of all recorded attacks worldwide, according to a new report from Unico in collaboration with Liminal. The study warns that criminal networks are using AI tools to create fake documents, manipulated videos, deepfakes, and synthetic identities faster and at a lower cost than in previous years.
In Mexico, fraud alerts are concentrated in states with the highest economic activity and digital identity verification volumes. Mexico City leads with 13.85% of the country’s alerts, followed by Veracruz (9.26%), the State of Mexico (8.66%), Jalisco (6.20%), and Chihuahua (5.95%).
By contrast, Baja California Sur, Zacatecas, Chiapas, Quintana Roo, and Aguascalientes each account for less than 1% of national alerts.
The report cites Liminal’s Global Fraud and Risk Outlook 2026, which projects that fraud losses at financial institutions will increase by more than 120% by 2030, rising from approximately $25 billion in 2025 to $55.3 billion by the end of the decade. A separate TechRadar study estimates that global financial fraud losses already exceed $400 billion annually.
Medium-complexity forgeries — such as images and videos manipulated through injection techniques — represent 45.8% of fraud attempts. Advanced attacks combining deepfakes, physical alterations, and other evasion methods account for 23.3% of the total.
When successful, deepfake-based fraud causes average losses of more than $280,000 per organization, with nearly 20% of cases exceeding $500,000, according to the Deepfake Impact Survey 2025 by IRONSCALES.
Fernando Paulín, general director of Unico Mexico, noted that the main shift is not just technological sophistication but the speed at which criminal groups share tools and reuse identities to target multiple companies. Unico’s intelligence network detected one fraudulent profile linked to 949 different identities that attempted transactions at 30 companies, highlighting the ability of these networks to replicate attacks at scale.
