The AI companies are feeding the press sensationalized stories. The AIs did not "break out of" anything. They were instructed to exploit vulnerabilities, and that's what they did. The problem was human error- they were supposed to be in air-gapped environments, but were given access to the Internet.
