AI-Generated Deepfake Scam — Video Calls, Photos & Voice Manipulation

Deepfake technology powered by artificial intelligence allows scammers to create convincing fake video calls, cloned voices, and manipulated photos. In one high-profile case, a victim lost S$4.9 million after being deceived by a deepfake video call appearing to show Singapore's Prime Minister. This technology is now accessible enough for everyday scammers to use against ordinary Malaysians.

As AI tools become cheaper and more convincing, deepfake scams are expected to grow rapidly across Southeast Asia.

Scammers use three main deepfake techniques. Voice cloning requires only 3-5 seconds of audio to create a convincing replica. The scammer obtains voice samples from social media videos, voice notes, or brief phone calls, then uses AI to generate real-time synthetic speech that sounds like the victim's family member or boss.

Video deepfakes are more complex but increasingly feasible. Scammers create fake video calls that show the face and mannerisms of a known person. These are used in business email compromise variants where the "CEO" calls an employee via video to authorize an urgent transfer.

Photo manipulation is the simplest form. Scammers generate fake identity documents, fake proof of investment returns, or fake images of celebrities endorsing investment schemes.

How realistic are deepfake video calls?

Current AI technology can produce video deepfakes that are convincing on mobile screens, especially during brief calls. However, they often show artifacts around facial edges, have slight lip-sync delays, and struggle with unusual head movements.

How can I protect my family from deepfake scams?

Establish a family code word for verifying identity during emergency calls. Limit the amount of personal video and audio you share publicly on social media, as these can be used to train deepfake models.