'Not a theoretical risk,' feds warn as attackers use AI-made code to hack critical infrastructure controllers

'It is an active threat'