<ul data-eligibleForWebStory="false"><li>AI models could unlearn how to imitate specific voices to combat audio deepfakes.</li><li>Machine unlearning aims to prevent voice identity fraud by making AI 'forget' particular voices.</li><li>This technique requires replacing data with randomness to ensure complete 'forgetfulness' by model.</li><li>Results show effective 'forgetting' of voices, but with some compromise on mimicking permitted voices.</li></ul>

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

Discover more