The key to spotting a deepfake? Look beyond the audio or…

In July 2024, a 69-year-old retiree named Martin Ng encountered a deepfake video on YouTube featuring Senior Minister Lee Hsien Loong. The video, which appeared to be an advertisement for a cryptocurrency investment scheme, used AI-generated content to mimic the political figure. While Ng initially found the video convincing, he identified it as a scam due to the voice, which he described as robotic with unnatural intonations.

What happened

The incident involved the unauthorized use of a high-profile political figure to lend credibility to a fraudulent financial scheme. Scammers utilized deepfake technology to create a video of the Senior Minister promoting investment opportunities. Although Ng did not fall for the scam, he noted that the technology has become increasingly realistic over time.

Experts from the Cyber Security Agency of Singapore (CSA) emphasize that as AI-generated content improves, visual and audio cues—such as lip-syncing issues or skin complexion—are becoming less reliable indicators of fraud. Instead, the CSA recommends focusing on the context of the message. Warning signs include promises of guaranteed high returns, requests for sensitive financial information, or high-pressure tactics designed to force immediate action.

This incident highlights a broader trend identified by Interpol, which reported a 600 per cent rise in deepfake discussions on cybercriminal forums between February and June 2024. Scammers frequently combine deepfake technology with traditional social engineering tactics, such as leveraging perceived authority to build trust. In a separate 2025 case, a finance director was targeted in a video conference featuring deepfakes of senior executives, nearly resulting in a loss of US$500,000. These cases demonstrate that relying on official channels and verifying information through trusted sources remains the most effective defense against increasingly sophisticated AI-enabled fraud.