As criminals harness AI to mimic voices from fleeting audio samples, experts warn of increasing scams that exploit social engineering. Practical steps include slowing down transactions and verifying identities through alternative channels to safeguard personal and corporate finances.
Artificial intelligence has made a familiar voice less trustworthy than it once was. Criminals can now mimic a relative, colleague or executive well enough to pressure victims into sending money or sharing sensitive information, and the tactic is spreading fast. The danger lies not only in the technology, but in how quickly it can be deployed using scraps of publicly available audio from social media clips, voicemail greetings and recorded calls.
Research cited by cybersecurity groups suggests that voice cloning can be produced from only a few seconds of sound, making everyday online activity a potential source of raw material for fraudsters. Mint reported that scammers commonly harvest clips from social platforms and voicemail messages, while Silent Security said modern tools can generate a convincing imitation from as little as three seconds of audio. That is enough to turn a public video or a casual greeting into a weapon.
The most damaging version of the scam is the fake emergency call. A cloned voice may claim to be a child, grandchild or spouse in urgent trouble and demand immediate payment through methods that are hard to reverse, such as wire transfers, cryptocurrency or gift cards. Scam Adviser and Mint both said the safest response is to hang up, call the person back on a known number and use a family safe word or other private check that an impostor could not know.
Businesses face the same threat in a corporate setting. Fraudsters have used cloned voices and deepfake video to impersonate executives, vendors and co-workers in attempts to divert payments or reset account access. Security researchers interviewed across the related reports say the pressure tactics are often the same: urgency, secrecy and a demand for action before anyone has time to verify the request. For organisations, that means one call should never be enough to authorise a transfer or alter payment instructions.
The practical defence is to slow everything down. Officials and fraud experts advise people not to rely on voice recognition alone, not to trust a request that arrives with unusual urgency and not to send money to settle a supposed crisis without checking through another channel. Limiting public audio, tightening social media privacy settings and simplifying voicemail greetings can also reduce exposure. If money has already been sent or personal information shared, victims should contact their bank immediately and report the fraud to the relevant authorities.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
