AI scams escalate! Cryptocurrency ATM fraud losses may reach $333 million in 2025, with US complaints soaring by 33%

On March 12, according to Cointelegraph, cybersecurity firm CertiK’s latest report shows that as artificial intelligence and deepfake technology are widely exploited by scam networks, the scale of cryptocurrency ATM scams in the United States is rapidly expanding. The report predicts that by 2025, losses from such scams could reach approximately $333 million, with related complaints increasing by 33% year-over-year, making it one of the fastest-growing types of financial crime.

The report points out that cryptocurrency ATMs (also known as crypto self-service terminals) are being heavily exploited by criminal groups due to their features of “quick exchange” and “low authentication.” Users can typically convert cash into digital assets in about five minutes, and some machines require minimal identity verification, making them convenient channels for scammers to transfer funds and launder money. Statistics show that the U.S. currently accounts for about 78% of the approximately 45,000 cryptocurrency ATMs worldwide.

Data from the Federal Bureau of Investigation (FBI) indicates that from January to November 2025, the U.S. received over 12,000 scam complaints related to cryptocurrency ATMs, a roughly one-third increase from the previous year. Security researchers believe that scam groups are using social engineering to induce victims to actively transfer funds on the terminals, facilitating the movement of money.

The report also mentions that about 86% of the losses among victims involve users over 60 years old. The study suggests that seniors generally have limited understanding of how crypto assets work and are more susceptible to phone scams or impersonation of government officials. Additionally, some younger users have fallen victim to so-called “romance investment scams,” commonly referred to as “pig butchering” scams in the crypto space.

Beyond emotional scams, fraudsters frequently use tactics such as impersonating government officials, technical support scams, “grandparent scams,” and fake fund recovery services. Most of these scams rely on psychological manipulation rather than technical vulnerabilities, establishing trust to force victims to perform transfers at ATMs.

CertiK further notes that AI tools are significantly increasing scam efficiency. AI can scrape social media information and generate highly personalized scam scripts, even using real-time deepfake videos or voice mimicking familiar people to victims, thereby boosting success rates. The report estimates that AI-driven social engineering scams could be 4.5 times more profitable than traditional methods.

Meanwhile, U.S. regulators are beginning to pay attention to these risks. Senator Cynthia Lummis has stated that future legislation on the crypto market structure should both combat scams and not hinder technological innovation. Another senator, Dick Durbin, has proposed the “Cryptocurrency ATM Fraud Prevention Act,” aimed at adding security protections for users of crypto self-service terminals.