Artificial intelligence is automating parts of fraud operations, making scams cheaper to run and more difficult to detect. Rather than eliminating scammers, AI is giving criminal enterprises new efficiency tools that reduce their reliance on large teams while expanding their reach.
According to research by Chainalysis, scam operations with documented links to AI vendors generated average on-chain revenue of $3.2 million per operation, compared with $719,000 for operations without such links. The difference reflects why criminals have adopted automation: AI can maintain fake identities, generate convincing communications, and manage multiple conversations simultaneously at a fraction of the previous cost.
How AI Reduces Fraud Operations Costs
Traditional confidence fraud requires sustained human effort. A fake investment adviser must answer questions, a romance scammer must remember previous conversations, and an impersonator must maintain a believable persona. AI systems can now perform these functions at scale, allowing one operator to manage work that previously required multiple people.
According to FATF president Giles Thomson, AI could enable "one or two people in a basement with a very big server" to conduct operations that once required much larger teams. The technology can generate convincing documents, images, voices, and identities while maintaining dozens of conversations in multiple languages simultaneously.
Documented Impact on Fraud Losses
The FBI's 2025 Internet Crime Complaint Center report recorded 22,364 complaints containing AI-related information and approximately $893.3 million in adjusted losses. Cases include fake romantic identities, business impersonation, and other fraud schemes built on impersonating real people.
Anthropic documented a direct example in its August 2025 misuse report, describing an actor using Claude Code in an extortion campaign targeting at least 17 organizations. The AI model assisted with technical work, information analysis, and preparing extortion demands that sometimes exceeded $500,000.
Labor-Intensive Operations Continue Alongside Automation
Scam compounds operating in Southeast Asia have not disappeared with automation. An Amnesty International investigation in 2025 documented at least 53 sites in Cambodia and interviewed 58 survivors from eight nationalities, finding evidence of trafficking, forced labor, confinement, and violence.
A FinCEN analysis from September 2025 found large transnational criminal organizations operating alongside AI-enabled services and other criminal infrastructure. Automation doesn't necessarily shrink the industry; it can increase output from existing workforces by allowing operations to attempt more fraud with the same or fewer people.
The Defense Challenge
Some organizations have deployed their own AI defenses. O2 created Daisy, an AI bot designed to keep phone scammers talking for extended periods. In the company's account, Daisy responded to scammers more than 1,000 times, with some calls lasting around 40 minutes.
However, such defenses become less effective if scammers also automate their side of interactions. A bot conversation between a fake grandmother and a scammer's automated system wastes neither human time nor generates meaningful losses for criminals.
Where Fraud Still Requires Physical Infrastructure
Payment systems and financial institutions remain harder to automate away. Criminals need bank accounts, exchanges, payment services, or mule networks to receive and move proceeds. They need victims to authorize transfers and institutions to process payments.
Interpol's 2026 global fraud assessment describes more than 1,500 transnational fraud cases involving $1.1 billion in reported losses, highlighting ongoing international efforts to stop payments after fraud has been detected.
As AI makes impersonation cheaper and more convincing, the burden of verification increasingly falls on ordinary people. Familiar voices carry less trust when they can be generated inexpensively, and video calls provide weaker evidence when faces can be synthesized. The result is that victims must do more investigation while criminals require less manual effort.
