Financial frauds are worth as much as Denmark's GDP. Interpol points the finger at AI
Global financial frauds have caused estimated losses of $442 billion in 2025, a figure approximately equivalent to Denmark's GDP. This is certified by the Global Financial Fraud Threat Assessment 2026 by Interpol, which assesses the overall global risk as high and forecasts that losses will increase significantly in the next three to five years. The international organization defines the scenario with a particularly descriptive phrase: "industrialization of fraud". AI, economic digital tools, and cross-border criminal collaboration have transformed online scams from almost "craft" activities into scalable infrastructures.
AI-driven fraud and fraud-as-a-service: the technological leap
The report documents a qualitative change in attack mechanisms. AI-enhanced fraud is already 4.5 times more profitable than traditional methods. Agent-based AI systems are now capable of independently planning and executing complete scam campaigns, from reconnaissance to ransom demands, without ongoing human intervention.
At the same time, the underground market has made tools accessible. Voice cloning, face-swapping, and instant translation are available starting from $50 per month through fraud-as-a-service marketplaces on the dark web. Kits sold in these channels offer tiered pricing, customer support, and plug-and-play components: with just a few tools, it's possible to produce convincing scans of fake identity documents in a matter of hours. Those who once had to build technical skills now simply purchase a subscription.
The human cost: compounds in Asia and the Yahoo Boys of Lagos
The UN estimates that at least 300,000 people are currently working in scam operations in Southeast Asia, in conditions of severe exploitation and abuse: a report from the UN High Commissioner for Human Rights from February 2026 documented torture, sexual abuse, forced abortions, and food deprivation in Myanmar, Cambodia, and Laos. The victims come from at least 66 countries, lured by false job advertisements. In Sihanoukville and similar cities, selling people into forced labor has become a common method to settle debts. The price of freedom from the compounds is typically more than $50,000.
Researchers Mark Bo, Ivan Franceschini, and Ling Li, in their 2025 book Scam, describe these settlements as real corporations: within them, scam companies, canteens, clinics, and brothels coexist. Workers manage multiple phones simultaneously from dawn to midnight; those who fail to meet performance targets face beatings.
A different profile of the phenomenon emerges from The Yahoo Boys, published on June 9 by Farrar, Straus and Giroux, where journalist Carlos Barragan recounts years spent in a poor suburb of Lagos alongside four fraudsters. The economic context that Barragan documents is essential to understanding the scale: a legitimate job in Lagos can pay 25,000 naira ($18) per month, while a 50-kilogram bag of rice costs 30,000, a price that has more than doubled since the start of the reporting. Among the fraudsters consulted, most estimated that between 60 and 80% of young men in Lagos are involved in some form of online fraud; an estimate impossible to independently verify but consistent with public statements from Nigerian authorities, including the director of the Economic and Financial Crimes Commission, who arrived at similar figures in 2023. Unlike the industrialized Asian compounds, the profiles that Barragan documents rely mainly on human ingenuity: their tool is social engineering, the ability to exploit loneliness and trust through text messages and social media.
At the end of April 2026, a joint operation between the FBI, the Chinese Ministry of Public Security, and the Dubai police led to raids on nine fraud centers in the United Arab Emirates, resulting in 276 arrests and the freezing of over $701 million in cryptocurrencies. The targets were pig-butchering schemes, fraud based on fictitious romantic relationships built to convince victims to invest in cryptocurrencies.
The knot that the Interpol report leaves open is both technical and economic: creating a convincing fraud costs less and requires less effort than detecting it. Biometric verification systems, forensic deepfake detection platforms, and AI-powered anti-fraud tools are attracting investments, but they start from a disadvantageous position by definition, as is always the case with security. With agent-based AI further lowering the entry threshold for attackers, the gap seems destined to widen before any countermeasure can have a conclusive, even partial, effect.