Crypto fraud is experiencing its most lucrative period in history. According to Chainalysis, fraudsters made more than $17 billion in 2025 alone. Researchers believe that one of the main factors driving this growth is artificial intelligence, which has dramatically reduced the cost of setting up fraudulent campaigns.

Over the past few years, artificial intelligence has been actively integrated into the cryptocurrency industry. Exchanges use it to detect suspicious transactions, developers use it to write code, and analysts use it to process massive amounts of data.

However, scammers were quick to adopt these same technologies.

According to the Chainalysis 2026 Crypto Crime Report and the TRM Labs 2026 Crypto Crime Report, generative AI has created virtually no new methods of fraud. Its primary role has turned out to be far more dangerous—it has automated existing schemes, reduced the cost of launching them, and made it possible to scale attacks with virtually no increase in costs.

That is precisely why analysts are increasingly talking not about the growth of individual criminal groups, but about the emergence of a full-fledged digital fraud industry.

---

Record volumes

According to a preliminary estimate by Chainalysis, cryptocurrency scams generated more than $17 billion for criminals in 2025. This is the highest figure in the company’s history.

TRM Labs cites an even more impressive figure. According to analysts' estimates, approximately $158 billion has passed through wallets linked to illegal activities.

The researchers do not claim that all of this growth is caused solely by artificial intelligence. However, both companies note the same trend: modern AI tools have significantly reduced the cost of setting up fraudulent transactions and made it possible to carry them out on a much larger scale.

---

Artificial intelligence has transformed the economics of fraud

The main change did not occur in the fraud schemes themselves.

Phishing websites, fake customer support services, fraudulent investment schemes, and spam emails existed long before ChatGPT was introduced.

The cost of producing them has changed.

Just a few years ago, launching a large-scale scam campaign required the involvement of developers, designers, translators, support agents, and mass-mailing specialists.

Today, generative models perform a significant portion of this work.

In just a few minutes, you can create a compelling text free of grammatical errors, translate it into dozens of languages, generate images of support staff, create a voice message, or produce a deepfake video.

What used to take weeks of preparation can often be completed in just a few hours today.

---

The figures that surprised the researchers

Chainalysis separately compared fraud schemes that use specialized AI tools with traditional operations.

The results were revealing.

Indicator	AI Schemes	Common Schemes
Average amount of funds	$3.2 million	$719,000
Daily Incoming Payments	35,1	3,89
Median daily income	nearly 9 times higher	—

The researchers emphasize that this refers only to cases where the use of AI could be confirmed through an analysis of blockchain transactions. Nevertheless, even this sample demonstrates just how significantly automation has transformed the efficiency of such operations.

---

The fastest-growing category

Chainalysis pays particular attention to impersonation scams —a type of fraud in which criminals pose as employees of banks, cryptocurrency exchanges, payment services, or government agencies.

According to the company, the volume of such transactions has increased by approximately 1, 400% over the past year .

The reason is obvious.

Modern generative models make it possible to create nearly perfect copies of websites, write error-free text, generate images of support staff, clone voices, and use deepfakes during video calls.

While users might have been put off in the past by poor translations or subpar design, such issues are becoming increasingly rare today.

It didn't take long to find evidence supporting this hypothesis. During their investigation, analysts at Chainalysis discovered the Darcula platform—a service that clearly demonstrates how modern technology is turning fraud into a fully-fledged commercial product

---

When Fraud Becomes a Service

One of the most telling examples of the new model is the Darcula platform, which Chainalysis analyzes in detail in its report.

Darcula does not seek out victims or steal money on its own.

It provides criminals with a ready-made infrastructure for carrying out phishing attacks.

Researchers discovered a control panel, tools for mass SMS campaigns, page templates for well-known companies, and a system for managing fraudulent campaigns.

It was through platforms like these that large-scale SMS attacks were carried out, masquerading as notifications from banks, delivery services, and the E-ZPass toll payment system.

According to Chainalysis, Darcula was developed by the Smishing Triad, a group that turned SMS phishing into a commercial product.

Darcula became one of the first examples of digital fraud operating on the model of a typical cloud service.

---

Crime-as-a-Service

Darcula is by no means the only example.

Chainalysis and TRM Labs note the rapid growth of the " Crime-as-a-Service" model, in which criminal groups specialize in specific stages of fraudulent operations.

Some people create phishing platforms.

Others are developing deepfakes.

Third parties sell voice cloning tools.

The fourth group handles mass mailings.

The fifth group helps launder the stolen funds.

This division of labor makes the market significantly more efficient and allows new models to be scaled up quickly.

---

Why Is This Dangerous?

The main problem is not that artificial intelligence has learned to persuade people more effectively.

The danger lies elsewhere.

AI has dramatically lowered the technical barrier to entry.

Whereas in the past, organizing a large-scale fraud campaign required expertise, a team, and significant financial investment, today many of these tasks are automated or available through off-the-shelf services.

The cheaper it becomes to launch such operations, the more people have the opportunity to engage in fraud.

It is precisely this economic impact that analysts consider to be one of the main consequences of the spread of generative artificial intelligence.

Conclusion by KLYO

Just recently, we discussed how algorithms and artificial intelligence trade faster than humans, compete with one another, and account for a significant portion of the activity on blockchains.

Now we see the other side of those same technologies.

The very same neural networks that help write code, analyze the market, and automate work are already being used to create phishing sites, deepfakes, and large-scale fraud campaigns.

Technology in and of itself is neither good nor bad.

It all depends on whose hands they end up in.