The risk isn’t just misuse, it’s loss of control. We already have evidence that some AI systems are behaving in ways their creators didn’t anticipate and can’t fully contain. One system became “so hungry for computing power” that it attacked its own network, seizing resources

Video