MMOOZZGG.com

Glossary

What is AI fraud?

In short

AI fraud is deception in which criminals use neural networks to fake voices, video, photos and convincing emails. Their aim is to get money or data.

Security and law · AI fraud

MMOOZZGG.com editorsChecked: 1 min read

For a child

Imagine someone puts on a mask that looks just like your friend’s face and asks for your backpack. Today a computer can make masks like that for a voice and a video. So there is an important rule: if someone asks for money or a secret, even in a voice you know, tell an adult first. Then check together whether it is really your friend.

In plain words

Neural networks can write correct emails, copy a voice and make a video with the face of a real person. Fraudsters use this to call in the name of a relative, a manager or a bank and ask for an urgent money transfer. In the past, spelling mistakes or a strange voice often gave a scam away; now these signs are disappearing. The main defense is the same as before: don’t decide about money in a hurry, call back on a number you know, and agree on a secret word that only your family knows.

For business

For companies, the main threat is fake calls and video meetings in the name of a manager who asks to pay an invoice urgently. Another is emails from suppliers with changed bank details. Accounting, finance and assistants are the targets. Build the defense on process, not on the skill of spotting a fake by eye. Confirm any payment requested by voice or video over a second channel, check a change of bank details with a call to the number in the contract, and have two people approve large sums. Train staff on real scenarios, and limit public voice and video recordings of managers where you can. Write these rules into a company procedure before an incident happens.

For a specialist

AI-enabled fraud combines social engineering with synthetic content. The FBI warning IC3 of 3 December 2024 names four channels: generated text for phishing and fake profiles, images for fake documents, voice cloning for calls in the name of relatives and for access to accounts, and video to pose as executives in online calls. The technology behind it is text-to-speech tuned to a voice, deepfake video and language models. The countermeasures are out-of-band verification of payment requests, code phrases, and multi-factor authentication where voice biometrics is not the only factor. Detection of synthetic content (liveness detection, artifact analysis) and generator watermarks also help. The FBI also advises limiting the photos and voice recordings you publish, and checking a caller by calling back on a known number.

Example

An accountant gets a call in the voice of the CFO, who asks her to pay a new supplier’s invoice urgently. Following the procedure, she hangs up and calls the CFO back on a known number, and the fraud comes to light.

Related terms

Questions and answers

How do you recognize AI fraud? Be careful when you see urgency, a request for money or data, and an unusual channel, even if the voice and face are familiar. It is safer not to guess whether something is fake, but to check the request through another channel.
How can a company protect itself from calls with a fake voice of a manager? Set a rule: a payment requested by voice is confirmed through a second channel and approved by two employees. Train accounting staff and assistants, and limit public recordings of managers’ voices.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us