Dr Simon Nieder warns that AI disasters will likely avoid a single, dramatic event like Hiroshima. Instead, risks emerge from incremental failures, such as AI-assisted pathogen design or infrastructure vulnerabilities. This shift in threat modeling suggests practitioners must prioritize continuous monitoring over searching for a specific trigger.