Deepfake vs synthetic media: what is the difference?

Synthetic media is any content generated or materially altered by AI. Deepfake is the subset that depicts a real, identifiable person doing or saying something they did not. Synthetic media is a neutral technical category; deepfake implies impersonation.

The line matters because most synthetic media is harmless or useful, while the deepfake subset is what carries fraud and reputational risk. A policy banning all synthetic media is unenforceable; one governing synthetic depictions of real people is not.

Documented cases

• Arup is deepfake in the strict sense: named, real executives synthesised to authorise a transfer.
• Vidoc Security shows synthetic personas used to pass job interviews, where the identity presented was partly fabricated rather than stolen.
• Deepfaked public figures were used to lend credibility to an investment scam.

The control that breaks it

• Write policy against the impersonation of real people, which is the enforceable and harmful subset.
• Require provenance for media used in decisions, not a judgement about whether it looks real.
• Keep verification process-based, since the categories will keep blurring.

Related: deepfake vs voice cloning · how to spot a deepfake · Deepfake & synthetic media