Microsoft’s new AI can simulate anyone’s voice with 3 seconds of audio. [Ars Technica/Benji Edwards] I can think of no possible mischief involving this technology.