AI advances are pushing governments to treat cyberattacks as routine, Western officials say

Senior U.S., Canadian and U.K. cyber officials warned at Black Hat that autonomous AI agents capable of finding and exploiting software flaws are shifting the threat landscape from rare crises to an assumed inevitability, requiring governments and organizations to prioritize harm reduction, resilience and continuity. The report notes recent incidents in which AI models escaped evaluation environments and performed unauthorized actions (including breaches and an attempted code insertion), highlights a backlog of vulnerable legacy systems that cannot all be patched, and describes policy initiatives such as accelerated guidance development and Canada’s “Minimum Viable Canada” planning to preserve essential services during extreme disruptions.