What are deepfakes and how can I spot them?

A deepfake is a photo, video or audio clip that has been created or changed using AI to make a real person look or sound like they said or did something they never did. The name combines "deep learning" (a type of AI) and "fake".

How are deepfakes made?

An AI model is trained on photos, videos or voice recordings of a person. It learns the patterns of their face, expressions and voice, then generates new content that mimics them. Some tools today need only a few photos or a short voice clip.

Why do deepfakes matter?

Deepfakes have some legitimate uses, such as special effects in films or recreating the voice of someone who has lost the ability to speak. But they can also be misused to:

• Scam people: in May 2026, a businessman in Singapore was reported to have lost at least S$4.9 million after joining a video call in which the "government officials" he met, including the Prime Minister, were all deepfakes.
• Spread false information: for example, fake clips of politicians or of news events that never happened.
• Bully or harass: for example, fake images or videos of classmates made to embarrass or hurt them.

How to spot a deepfake: the 3A approach

Singapore's Cyber Security Agency recommends three steps:

1. Assess the message

• Who sent it, and is the source trustworthy?
• Does it make sense? Would this person really say or do this?
• Is it urgent, asking for money or personal details, or pushing you to download something? These are major warning signs.

2. Analyse the audio and visuals

• Unnatural facial expressions or blinking
• Blurring around the edges of the face or hair
• Lighting or skin tone that does not match the background
• Lip movements that do not match the words, or a voice that sounds flat or robotic

3. Authenticate the content

• Check official websites or trusted news outlets to see if the story is real.
• Use a reverse image search to find where an image first appeared.
• Contact the person directly through a channel you know is genuine.

Why your eyes alone are not enough

AI tools are improving quickly, and many deepfakes no longer show obvious glitches. A video that looks perfect can still be fake. That is why checking the source (Assess) and verifying with trusted sources (Authenticate) matter more than spotting visual flaws.

What should you do if you see one?

• Do not share it, even as a joke. Sharing spreads the harm.
• Report it on the platform where you saw it.
• Tell a trusted adult, such as a parent or teacher, especially if it involves you or someone you know.
• If you think it is a scam, do not send money or personal details. Contact official anti-scam helpline or the police to check and report it.

Key takeaway: Deepfakes use AI to make fake content look real. Do not rely on your eyes alone: question the source, check with trusted sources, and pause before you believe or share.