Key facts
- AI has made crypto fraud cheaper, more personalized, and more convincing, particularly through impersonation.
- The FBI reported $20.9 billion in cybercrime losses in 2025, with crypto being a frequent payment method.
- Chainalysis found AI-powered crypto scams were 4.5 times more profitable than those without.
- Traditional financial controls like dual authorization, out-of-band verification, and independent reconciliation are essential defenses.
- Programmable smart accounts and automated monitoring can create programmatic security guardrails.
Artificial intelligence is transforming the landscape of financial fraud, making impersonation tactics cheaper and more convincing, according to a recent analysis for financial advisors.
The FBI's Internet Crime Complaint Center reported a record $20.9 billion in cybercrime losses in 2025, with cryptocurrency being the most common payment channel. Chainalysis estimates that approximately $17 billion flowed to crypto scams during the same period, noting that operations leveraging AI tools were roughly 4.5 times more profitable than those without. The average payment in crypto scams more than tripled year-over-year to $2,764.
Instead of focusing solely on detecting sophisticated AI-generated fakes, advisors are urged to reinforce traditional financial controls. These include dual authorization for all asset movements, requiring two independent approvers to prevent a single impersonated individual from authorizing transactions. Out-of-band verification, confirming instructions through a separate, pre-agreed channel, is also critical, especially when faced with urgent requests.
Independent reconciliation of client holdings against verifiable on-chain data, performed by someone not involved in initiating transactions, is another key control. Advisors should also conduct thorough due diligence on custodians, reviewing their SOC reports and asset-segregation practices. The digital-asset accounting standard ASU 2023-08 provides further disclosure opportunities for verification.
Experts suggest that AI has not introduced entirely new fraud categories but has significantly reduced the cost of executing existing ones. This shift places a greater emphasis on robust, verified processes. For money managers, transitioning to programmable smart accounts, such as those compliant with ERC-4337 or EIP-7702, can enable automated security guardrails at the account level, including continuous monitoring for suspicious activity and blocking risky approvals before funds can be moved. While AI can assist advisors by flagging unusual behavior, granting it direct, unmitigated wallet permissions is seen as a major vulnerability.