Google reports that attackers have used an autonomous artificial intelligence framework called exploit_agent to turn security patches into working exploits with no human in the loop. The group also found AI being used to automate live intrusions, run massive influence operations, and deploy backdoors.