Can AI fool a bank's voice ID or identity checks?
PartlyPartly. Voice-only checks have been beaten with cloned voices and deepfaked IDs are getting past some document checks, but layered checks (app, PIN, device) still hold up much better.
First shown: Feb 23, 2023. Vice journalist Joseph Cox used an AI clone of his own voice, made with ElevenLabs' free service, to pass Lloyds Bank's Voice ID and reach his account. (Vice)
What it means
Banks check identity in several ways: your voice on the phone, a photo of your ID, a selfie or live video, a PIN, and the phone or app you use. AI attacks the first three, which rely on what you look or sound like.
In February 2023 a journalist played a cloned version of his own voice to Lloyds Bank's phone line and was let in. It took several failed attempts and also needed his date of birth. In July 2025 OpenAI's Sam Altman told a Federal Reserve conference that some financial institutions still accept a voiceprint as authentication and that AI has 'fully defeated that'.
Document and face checks are under pressure too. US regulator FinCEN said in November 2024 that, beginning in 2023, it had seen a rise in suspicious-activity reports describing deepfake media, most often fake identity documents used to get around identity verification. Hong Kong police said in 2024 that deepfakes had beaten facial recognition at least 20 times in loan and bank-account applications.
Checks that do not depend on your face or voice, such as a code from your banking app, a hardware key or a PIN, are not defeated by deepfakes on their own. The weak point is any single check that relies on how you sound or look.
The evidence ladder
Every capability on this site is graded on the same five steps, from a single demonstration to real-world consequences.
| Step | Status | When | Evidence |
|---|---|---|---|
| Demonstrated shown at least once | yes | Feb 23, 2023 | A journalist's cloned voice passed Lloyds Bank's Voice ID after several failed attempts; he also had to enter his date of birth. (Vice) |
| Reproduced independently repeated | yes | Feb 2, 2024 | Hong Kong police said deepfakes had been used at least 20 times to defeat facial recognition in loan and bank-account applications made with stolen ID cards. (CNN) |
| Accessible available to ordinary people, or criminals | yes | Nov 13, 2024 | FinCEN said it had seen more suspicious-activity reports of deepfake media since 2023, most often fake identity documents used to get around identity checks; the Lloyds test used a free consumer tool. (FinCEN) |
| Reliable works most of the time for a defined task | partly | Jul 22, 2025 | Altman said AI has 'fully defeated' voiceprint authentication, but that is a claim rather than a measurement; the documented bank test needed repeated tries and personal data. (PYMNTS) |
| Consequences real-world use or harm documented | yes | Nov 13, 2024 | FinCEN issued alert FIN-2024-Alert004 on deepfake fraud against financial institutions, listing red flags such as a customer using a third-party webcam plugin during a live verification check. (FinCEN alert) |
From the record
- A UK journalist used AI-generated audio of his own voice to get past Lloyds Bank's Voice ID. (AI Incident Database)
- Alongside the Arup case, Hong Kong police reported deepfakes beating facial recognition in at least 20 loan and bank-account applications. The case file → (South China Morning Post)
- Dutch fraudsters used AI-generated calls posing as bank helpdesks to trick about 170 customers, rather than fooling the bank itself. The case file → (NL Times)
What it means for you
If your bank lets your voice or a selfie alone unlock your account, that lock is weaker than it was. Banks are adding layers, but the strongest protection is a check that relies on something you have or know, not how you sound.
What to do
- Switch off voice-as-passwordAsk your bank to disable voice ID on your account and use app approval or a PIN instead.
- Turn on app or key-based sign-inUse your banking app's approval prompts or a hardware security key where offered.
- Guard your date of birth and detailsVoice and ID attacks usually also need personal details; share less of them publicly.
- Remember banks call you, not the reverseA caller who sounds like your bank and pressures you to move money or install an app is a red flag; hang up and call the number on your card.
Questions people ask
Has AI actually fooled a bank's voice recognition?
Yes. In 2023 a journalist used a cloned voice to get into his own Lloyds Bank account, after several failed attempts and using his date of birth.
Is voice ID still safe to use?
It is the weakest option where alternatives exist. OpenAI's CEO told a Federal Reserve conference in July 2025 that AI has 'fully defeated' voiceprint authentication.
What are banks doing about deepfakes?
US regulator FinCEN issued an alert in November 2024 telling banks what red flags to watch for, such as a customer using a third-party webcam plugin during a live check.
Related
- AI voice clone scam call: what to do right now →
- AI scam texts: how to spot them and what to do (2026) →
- How to set up an AI chatbot for an older parent, safely →
Sources
- Vice: How I broke into a bank account with an AI-generated voice
- AI Incident Database, incident 485
- FinCEN alert on deepfake fraud (Nov 2024)
- FinCEN alert FIN-2024-Alert004 (PDF)
- PYMNTS on Altman's voiceprint warning (July 2025)
- CNN on Hong Kong deepfake fraud
Spot an error or something out of date? Tell us; corrections are made and logged. Written with AI assistance from public sources; every fact checked and approved by Colin Balfe. General information, not legal, medical or financial advice.