Synthetic media that looks real
A deepfake is a fake image, video, or audio recording created using deep learning — the name combines the two words. It can show a real person saying something they never said, or depict an event that never happened.
Why it is a problem
- Fraud: transfer scams using a cloned voice or face of a family member or a manager
- Disinformation: fabricated speeches or disaster footage used to manipulate opinion
- Defamation: non-consensual fake images and video that violate personal rights
- Erosion of evidence: genuine footage now gets dismissed as “probably AI”
Video generation has lowered the barrier so far that convincing fakes no longer require expertise, which is what makes this serious. Investment scams using fabricated celebrity endorsements have already caused real losses.
Can you spot them?
People used to point to unnatural hands or infrequent blinking. Those artifacts have largely disappeared, and visual detection is increasingly unreliable. Judging by provenance rather than by appearance is the more durable approach.
In practice: confirm through official announcements or several credible outlets; verify anything involving money or personal information through a separate channel; and pause before resharing. Standards for embedding provenance signals in AI-generated content are also progressing.
The technology itself is neutral
The same methods are used for film dubbing, de-aging actors, and restoring the voice of someone who has lost it to illness. It is the standard example of a technology being neutral while its use is not, which is why it recurs in discussions of AI literacy.
Assuming that convincing fakes can exist is becoming basic equipment for handling information at all.