Scammers stole an estimated $17 billion worth of cryptocurrency in 2025, according to a new report from Chainalysis. Notably, the report found that AI-assisted scams stole 4.5 times more money than scams that didn’t leverage AI.
“Our analysis reveals that, on average, scams with on-chain links to AI vendors extract $3.2 million per operation compared to $719,000 for those without an on-chain link — 4.5 times more revenue per scam,” the researchers write.
“These AI-related operations also demonstrate significantly greater time-weighed efficiency....These metrics suggest both higher operational efficiency and potentially broader victim reach. The increased transaction volume indicates that AI is enabling scammers to reach and manage more victims simultaneously, a trend consistent with the industrialization of fraud we’ve been tracking. In contrast, the increased scam volume suggests that AI is likewise making scams more persuasive.”
These scams are also driven by sophisticated phishing kits that allow unskilled threat actors to launch industrial-scale fraud operations.
“Many of these campaigns have a social media angle, given that such platforms provide access to millions of users, and are thus prime targets for sending automated messages,” the report says.
“In such cases, scammers may buy bulk social media profiles and use SMS and phishing kits to communicate. The material impact of this large-scale industrialization cannot be understated. Scams leveraging these phishing kits are 688 times more effective in dollar terms and four times more effective in average transaction size than regular scams. Scams that buy bulk social media accounts are likewise 238 times more effective in dollar terms and two times more effective in average transaction value compared to regular scams.”
KnowBe4 empowers your workforce to make smarter security decisions every day. Over 70,000 organizations worldwide trust the KnowBe4 HRM+ platform to strengthen their security culture and reduce human risk.
Chainalysis has the story.
Here's how it works:
