AI can now generate speech that sounds natural and can mimic a specific person's voice from a short sample. This powers helpful uses, such as restoring a voice lost to illness or narrating content, but the same ability enables impersonation and scams.
Because a convincing fake voice can deceive people and bypass voice-based checks, attention has turned to consent and safeguards: permission to clone a voice, ways to mark audio as synthetic, and caution about trusting a voice alone. The technology is impressive and useful, but it has made clear that a familiar voice is no longer proof of who is speaking.