Safety & ethics
A deepfake is a fake image, video, or audio clip made with AI that shows a real person doing or saying something they never did, often convincingly.
The word combines 'deep learning' and 'fake.' It appeared in 2017 when a Reddit user of that name posted videos with celebrity faces swapped onto other people's bodies. Since then the tools have become cheap and easy, and the quality has jumped from clearly fake to hard to detect.
Deepfakes are made with generative models: GANs and, more recently, diffusion and voice-cloning systems. Given a few photos or seconds of audio, a model can learn a person's face or voice and reproduce it saying anything. Video, lip movement, and tone can all be generated together.
Harms include scams using a cloned relative's voice, fake political speeches, and explicit images made without consent. Defenses include detection tools, invisible watermarks on AI output, and laws in many countries as of 2026. The safest habit is simple: verify surprising media through a second channel.
A family gets a panicked phone call in their son's exact voice asking for money; the voice was cloned from clips on his social media.