AI text-to-speech programs could “unlearn” how to imitate certain people

Researchers are developing a method called "machine unlearning" that can train AI text-to-speech models to forget how to imitate specific voices. This advancement is seen as a crucial defense against the growing threat of audio deepfakes, which are used for fraudulent activities and scams. The technology aims to prevent malicious actors from exploiting AI's ability to convincingly replicate human speech for harmful purposes.

This development is significant because it offers a potential technical solution to combat the misuse of advanced AI voice-cloning technology for criminal activities like fraud and scams.