US Banks Warn Generative AI Is Industrialising Scams

Generative artificial intelligence is allowing criminals to launch scams faster, more cheaply and at much greater scale, the American Bankers Association has warned.

In testimony to the US Senate Special Committee on Aging, ABA risk, fraud and cybersecurity executive Paul Benda said AI was making long-established fraud methods more effective, Public reporting indicates that .

Old scams gain new reach

AI can improve phishing messages, imitate voices and create personalised scripts from information available online. Automation also lets attackers contact far more victims while continuously testing which approaches work.

Banks should strengthen behavioural monitoring and encourage customers to verify urgent payment requests through a trusted, separate channel. Warning messages are most effective when tailored to the transaction context rather than presented as generic text.

Financial institutions also need rapid feedback loops between fraud operations, customer service and transaction monitoring. Generative AI changes the scale and quality of deception, but coordinated data and human intervention can still interrupt the payment before funds are lost.

Older customers may face particular exposure

Criminals frequently combine urgency, authority and emotional pressure, impersonating relatives, banks, government agencies or investment professionals. AI makes it easier to personalise these stories and reproduce a familiar voice, reducing the warning signs a victim might previously have noticed.

Firms should design interventions around the customer’s situation. A short delay, a call from trained staff or a question about how the beneficiary was introduced can interrupt manipulation. Banks should also track which scam scripts, channels and beneficiary accounts recur so controls improve as criminals change tactics.

Indicators of AI-enabled deception

• Highly personalised contact combined with urgent payment pressure.
• A familiar voice that refuses independent verification.
• Rapid changes in the beneficiary or stated reason for payment.
• Several customers receiving near-identical investment or impersonation scripts.

Next focus: Banks should measure which interventions actually stop AI-assisted scams and share those results. Controls that add friction without reducing losses may need to be redesigned around customer behaviour and beneficiary risk.