What it is
A deepfake is synthetic video or audio in which a real person says and does things that never happened. The quality has grown to the point where "I saw it with my own eyes" is no longer proof.
Where it came from
- The term is a blend of "deep learning" and "fake"; early fakes were crude and easy to spot.
- Real-time face and voice generation made live faked video calls possible.
- A landmark case: in 2024 the engineering firm Arup lost around $25M — an employee wired the money after a video call in which the "CFO" and the "colleagues" were all deepfakes.
Why it took off (in the news cycle)
- Accessibility: almost anyone can now make a convincing fake, with no studio.
- High stakes: fraud, disinformation and non-consensual content all at once.
- Every high-profile case gets amplified by the media and erodes basic trust in what we watch.
How to use it today
- Financial hygiene: transfers and "urgent requests from management" only go through a second channel with a code word, even if you saw the face.
- For content: mark your genuine video (provenance, signature) so it stands apart from fakes.
- Train your team: a video call is NOT proof of identity; verify independently.
What to watch out for
- The "liar's dividend": genuine video can now be dismissed as "that's a deepfake" — trust falls on both sides.
- Laws on non-consensual fakes and mandatory labelling are appearing — follow the regulation.
- Deepfake detectors are not 100% reliable; a verification process matters more than a magic button.