NIST, CISA and NSA publish incident response practices for rogue AI agents
Three US agencies have released ten voluntary practices to help frontier AI developers prepare for, detect and respond to incidents in which AI systems do not behave as intended.