A reporter says he used an AI-generated replica of a voice to bypass the voice verification system of the UK's Lloyds Bank and accessed an account's information
Banks in the U.S. and Europe tout voice ID as a secure way to log into your account. I proved it's possible to trick such systems with free or cheap AI-generated voices.
VICEJoseph Cox
Context & Ripple Effects
Voice cloning had already moved from a fraud risk in corporate payment authorization—a reported AI impersonation of a German executive—to an alleged $35 million theft using a cloned director’s voice. The Lloyds report brings that threat to a retail bank’s voice-verification channel rather than a human employee’s judgment.
It also anticipates a later test in which a columnist used AI tools to fool a bank voice-biometric system, suggesting that inexpensive synthetic audio challenges voiceprint checks themselves, not only call-center staff.
First-order effects
Lloyds Bank must treat voice verification as insufficient on its own for account-information access if the reporter’s test reflects its deployed controls.
Customers using Lloyds voice ID face a higher impersonation risk where an attacker can obtain or recreate enough of their speech.
Second-order effects
Banks and fintechs using voice biometrics are pushed toward additional authentication signals and tighter limits on what a voice-authenticated caller can view or change.
Voice-AI providers become part of financial-services fraud risk: tools that lower the cost of realistic replicas also raise the burden on banks’ fraud controls.
Third-order effects
If repeated bypasses hold up, voice biometrics will shift from a standalone login mechanism to one signal in layered, risk-based authentication for financial accounts.
The pattern supports a broader market for controls designed around synthetic identities and AI-mediated payment risk, rather than trust in a voiceprint as a durable identifier.
The trend: Consumer-bank authentication is moving away from treating a familiar voice as proof of identity as voice cloning becomes accessible enough to test and exploit verification systems.
New: we proved it could be done. I used an AI replica of my voice to break into my bank account. The AI tricked the bank into thinking it was talking to me. Could access my balances, transactions, etc. Shatters the idea that voice biometrics are foolproof https://www.vice.com/...…
To make this AI voice, I used just ~5 minutes of audio and uploaded to ElevenLabs, probably the most impressive and readily accessible AI-voice system at the moment (I read the GDPR into my mic). It's also being abused by 4chan to dox and harass https://www.vice.com/... https://t…
That's funny — someone was just telling me that scammers had been calling his elderly mother and had used their conversations to clone her voice in attempts to steal from her. https://twitter.com/...
speech is basically a type of video. A series of frames connected together by transitions. A lot of math, but yeah I would not bet on biometrics long term. AI will nullify this technique. Quantum computing will do the same to cryptography https://twitter.com/...
Good time to remind everyone that fraudsters used a cloned voice of a company director in a $35 million bank heist in 2021: https://www.forbes.com/... https://twitter.com/...
This story is wild. Readily-available deep faked voice clones can get past your bank's voice-print security. They take minutes to make. https://twitter.com/...
Biometrics are terrible for privacy but they're also just really really bad at all the security stuff that the industry wants them to do https://twitter.com/...
Journalist @josephfcox broke into a bank account using an AI-generated voice clone. The end of “my voice is my password” is nigh. Story: https://www.vice.com/... https://twitter.com/...
NEW: @josephfcox broke into his bank account using an AI-generated clone of his voice. This suggests new “voiceprint” tech deployed by many banks can be easily defeated https://www.vice.com/...
surely this is the worst thing AI can do and all the facial recognition technology isn't vulnerable too. and even if it is, it's not like a bunch of people uploaded millions of terabytes of selfies to AI a few months ago in the name of flattering AI photoshops, https://twitter.co…
There's a lot of AI moral panic going on right now, but this is an actual vulnerability. It takes a previous vulnerability — basically simjacking — to a whole new level. https://twitter.com/...
I've said it before and I'll say it again: voice “authentication” is easily thwarted and should never be used to verify identity. If your bank or other service uses voice “authentication” tools and you have an option to switch to multi-factor authentication — switch that now. htt…