A deepfake is a piece of synthetic media — usually video or audio — in which machine-learning models generate or alter a person's face, body, or voice to make them appear to say or do something they did not.
Context
During elections, deepfakes can be used to fabricate statements by candidates, forge "leaked" recordings, or discredit authentic media by casting doubt on everything (the "liar's dividend").
How investigators respond
- Look for provenance: where did the media first appear, and who published it?
- Check for artifacts: unnatural blinking, lip-sync errors, inconsistent lighting.
- Prefer verification of source and chain of custody over detector tools alone, which are imperfect.
#ai#video#synthetic-media#verification