Can AI autonomously wage cyberwar on critical infrastructure using ai-discovered and weaponized zero-day exploits ?

State and non-state actors are training AI models to probe industrial control systems, power grids, and defense networks for undetectable vulnerabilities. Unlike human hackers, these systems operate at machine speed, reverse-engineering protocols and chaining exploits in real time. Recent sandbox tests show AI agents identifying and weaponizing new zero-days within days—faster than agencies can patch or even catalog threats. The next major blackout or hospital shutdown may be authored by code rather than human intent.