AI-generated audio and banking fraud: how to verify a suspicious call or payment

A phone rings and a familiar voice asks you to move money now: a relative in trouble, an executive authorizing a transfer, or someone claiming to be from your bank. The voice can sound right and still be synthetic. That voice is a claim to verify, not proof of who is calling.

What the FBI says about AI-generated audio

The FBI's public warning (IC3, PSA241203) states that criminals exploit generative AI to commit fraud at a larger scale and to make it more believable. The FBI notes that generative AI can reduce the time and effort criminals spend deceiving a target, and can even correct the human errors that might otherwise warn you something is wrong.

Among the tactics the FBI lists, one bears directly on banking: criminals use AI-generated audio to impersonate public figures or personal relationships in order to elicit payments. The FBI also describes AI-generated text used in social engineering, spear phishing, romance, investment and other confidence schemes, and AI-generated images used for profile photos and identification documents. Creating or distributing synthetic content is not inherently illegal, the FBI adds, but it can facilitate fraud and extortion.

Why a familiar voice is a claim, not proof

The reflex to trust a voice you recognize — a family member, a manager, a "bank officer" — is exactly what a voice-cloning scam leans on. A voice that sounds correct does not confirm who is speaking, and a voice that sounds slightly off does not by itself prove a fake. Treat the caller's identity as something to confirm through a channel you control. The part to distrust most is the pressure to act immediately.

A verification routine you can use

These are steps this article suggests, not requirements published by the FBI or by any bank. Adapt them to your own bank's and employer's official procedures:

1. Hang up and call back on a number you already trust — the one printed on your card or on your bank's official site — never a number or link supplied during the call.
2. Slow the payment or account change down. Urgency is the lever; a genuine request survives a pause and a callback.
3. Agree a private verification detail in advance with close family or, for a finance team, an internal code word, so a caller can be checked without relying on the sound of the voice.
4. For business payments, require a second independent approver for any transfer or change of vendor bank details, confirmed through a separately opened official channel.
5. Write down exactly what was requested, who would receive the money, and why, before you act on anything.

The FBI's warning specifically links AI-generated audio impersonation to eliciting payments, which is why the money request — not the voice — deserves the scrutiny.

Where a detector fits, and its limits

If you have a recording — a voicemail or a saved clip — DeepFakeCheck can analyze that audio (as well as image, video, or text) and return a probabilistic risk signal. It can add one more signal to your decision.

Be clear about the limits. Automated detectors produce both false positives, flagging authentic audio, and false negatives, missing a synthetic clip. A score — high or low — cannot on its own prove who was on the call, that a payment is safe, or that a request is legitimate. A live call often leaves you no file to test at all, so the habit of calling back on a trusted number matters more than any single tool.

If money has already moved

If you have sent funds, or nearly did, treat it as suspected fraud rather than embarrassment. Stop any further payments, contact your bank or card issuer immediately, and preserve everything: the recording, the numbers, the messages, and the transaction details. The FBI directs victims of financial fraud to report to its Internet Crime Complaint Center at www.ic3.gov, and to include available transaction and communication details.

Sources

FBI Internet Crime Complaint Center (IC3), public service announcement, "Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud" — https://www.ic3.gov/PSA/2024/PSA241203