Voice cloning
Copying someone's voice from a short recording so a machine can say anything in it.
Voice cloning takes a sample of someone speaking and builds a model that can then say anything in that voice. What changed recently is how little it needs. A few seconds of clear audio is often enough, which means a voicemail greeting, a podcast appearance or a video you posted is plenty of raw material. The good uses are real: people who have lost the ability to speak can keep their own voice, and audiobooks and dubbing get cheaper to produce.
The trouble is that the same technology works just as well without permission. Scam calls using a cloned relative’s voice are now common enough to show up in national surveys, and performers whose voice is their livelihood have no clear legal claim to it in most countries, because the law protects recordings and names but rarely the voice itself. The practical defence against the scam version is low tech: agree a code word with your family that a caller would have to know.