AI-Powered Crypto Scams Hit Record $17B in 2025

The core finding of the Chainalysis report is the transformative impact of artificial intelligence on the economics of fraud. Scams with on-chain links to AI vendors generated an average of $3.2 million per operation, roughly 4.5 times more than scams without those links. This profitability stems from AI’s ability to solve the twin challenges of scale and believability. “On a time-weighted basis, you get faster scale and better believability,” explained Chainalysis Head of Research Eric Jardine. The data shows over 70% of AI-enabled scams exist in the top 50th percentile of transfer volume, meaning they are “getting bigger faster, and pulling in more money per transfer.”

The tools enabling this shift are face-swap software, deepfakes, and large language models, often sold by Chinese vendors through Telegram channels. Their most devastating application is in impersonation scams. “Once you move into these deepfake-type scenarios where people look, for all intents and purposes, like someone you know or a person of authority you’ve dealt with before, the believability goes up,” Jardine said. This heightened believability directly translates to higher victim payouts: the average scam payment rose to $2,764 in 2025, a 253% increase from $782 a year earlier.

Government impersonation has become particularly effective, with scams using deepfaked images of officials growing more than 1,400% in 2025. Criminals pose as workers from agencies, financial institutions, and crypto platforms to deceive victims. A stark example is the “Darcula” or “Smishing Triad” campaign, a Chinese group that targeted U.S. residents with fraudulent “E-ZPass” toll alerts, sending up to 330,000 texts daily using phishing kits that likely cost less than $500—demonstrating how low-cost, AI-enhanced infrastructure can yield massive returns.

Beyond impersonation, AI is supercharging more complex, long-term frauds. The report highlights “pig butchering” scams, where scammers build relationships—often romantic or financial—before persuading victims to transfer increasingly large sums. “You’re essentially trading off scale for believability,” Jardine noted, explaining why these relational scams have a higher average stolen amount than quick cons like fake YouTube giveaways. The danger was illustrated in December when a woman in San Jose, California, used ChatGPT to identify her new romantic partner as a pig-butchering scammer after losing nearly $1 million in cryptocurrency.

Concurrently, scammers are evolving their money-moving tactics. They are increasingly abandoning centralized exchanges for decentralized finance (DeFi) options like decentralized exchanges (DEXs), DeFi bridges, and protocols to launder their loot. Jardine explained this shift as part of a broader trend toward the decentralization of scam operations, leveraging the permissionless nature of DeFi tools to obscure fund flows. Furthermore, while basic automation often suffices for on-chain movement, advanced AI could soon be deployed “at that final point of reintegration” to create fake, KYC-compliant exchange accounts in bulk, helping scammers efficiently cash out into traditional currencies.

The financial devastation documented by Chainalysis is underpinned by a grim human reality. The ability to automate and scale fraud, particularly cashing out, helps sustain physical scam operations that have taken root in Southeast Asia. In recent years, so-called scam compounds have emerged across Myanmar and Cambodia, turning pig butchering into a massive industry fueled by human trafficking and forced labor. These operations, often run by Chinese organized crime networks, use specialized laundering channels to convert stolen crypto into luxury assets.

The scale of this integrated crisis was underscored in December when the U.S. Department of Justice moved to shut down domains linked to a major compound in Myanmar. “These cases demonstrate the scale of modern cryptocurrency scam operations and their increasing integration with traditional organized crime,” Chainalysis stated. The report concludes with a sobering reminder of the dual exploitation at play: “They also reveal the human cost of these schemes, which exploit both financial victims and the trafficked individuals forced to operate them, itself an unspeakable crime.” The record $17 billion in losses for 2025, therefore, represents not just a financial market report, but a measure of a deepening criminal ecosystem where AI, cryptocurrency, and human suffering are tragically intertwined.