What AI voice and deepfake scams look like today
The best-documented case is a company, not a family. In early 2024 an employee of the engineering firm Arup in Hong Kong joined a video call with what looked and sounded like the chief financial officer and colleagues, all generated, and made 15 transfers totalling HK$200 million, about US$25.6 million. Arup confirmed in May 2024 that fake voices and images were used, as ITV reported.
- Officials and executives. The FBI’s May 2025 announcement describes text and AI-generated voice messages impersonating senior US officials to gain access to accounts.
- Investment promotions. Deepfake videos of public figures advertise trading platforms; see the investment platforms guide.
- Suppliers and partners. A cloned voice confirms “new bank details” already sent by email.
- Robocalls. The FCC ruled in February 2024 that calls made with AI-generated voices are “artificial” under the TCPA, which makes them subject to US robocall rules (FCC news release).
Family emergency calls with a cloned voice are covered in the voice clone scams guide; this page is about requests that reach you at work or as an investor.
An AI scam call: what to do during the call
- 01Do not act on the call itself. Say you will call back, and end it.
- 02Call back on a number you already have, or one from the official website, never the number that called you.
- 03Ask a question only the real person can answer, or use the verification phrase your team agreed on. The FBI recommends a secret word or phrase for families in its December 2024 announcement; the same idea works for finance teams.
- 04Treat the request as the signal: urgency, secrecy, a new account, gift cards or crypto.
- 05If you can legally keep the voice message or the recording, keep it.
How to check a voice message or a video for AI
Our AI voice detector scores each segment of a recording (MP3, WAV, M4A, OGG, WebM or OPUS, up to 25 MB and 5 minutes) with an open-source anti-spoofing model, transcribes it, and flags scam patterns in the transcript: urgency, secrecy, money requests and changed bank details. The deepfake video detector checks one frame per second, the largest face when it is clearly visible, and the soundtrack, and marks suspicious moments on a timeline.
Both give an indication, not proof. Phone and messaging compression removes many traces, so a clean result on a short, compressed clip means little. The guide on how to tell if a voice is AI explains what still helps.
AI Voice Detector · Free plan
Check a voice message for cloning, with scam patterns flagged in the transcript.
Check the company behind the request
Most AI scams at work end with a payment to a company: a “new supplier”, a “law firm”, an “investment vehicle”. That company is where a scam usually fails, because a convincing voice does not create a registered business with a matching website.
The AI scam checker brings both halves together: the Business Check verifies the company (registries, people, domain age, regulator warning lists for financial firms), then the detectors check what it sent you. You can add up to three phone numbers that called you; we check validity, country and line type.
Any free account can run a Business Check (3 credits, out of 10 free credits a month) and see how many checks ran and how many warning signs came up. The full report, with the score, the company, the people and every source, is on the Business plan ($49 a month), with a manual review by our team within 48 hours when the AI cannot confirm everything.
Frequently asked questions
What is an AI voice cloning scam?
A scam in which a synthetic copy of a real person’s voice is used on a call or in a voice message to request money, codes or access. It targets families, employees and investors. Calling back on a number you already have stops most of them.
How can I tell if a call is an AI scam call?
You often cannot tell by ear. Judge the request instead: urgency, secrecy, a new payment method or new bank details. End the call and call back on a number you already hold. A recorded message can go through our voice detector as a second check.
Can deepfakes really fool a video call?
Yes. In the Arup case confirmed in May 2024, an employee made 15 transfers totalling about US$25.6 million after a video call in which the other participants were generated. Verification outside the call is what protects a company.
Are AI-generated robocalls legal in the US?
The FCC ruled in February 2024 that AI-generated voices count as “artificial” under the Telephone Consumer Protection Act, so robocalls using them fall under the same rules, including consent requirements.