AI has unearthed a severe underlying PHP vulnerability, PHP2Shell, which enables remote server manipulation and impacts versions stretching back more than ten years. Following the inadvertent revelation of this vulnerability, governments and vital institutions in various countries proactively shut down numerous related websites to minimize potential risks. Traditionally, such vulnerabilities were often disregarded because exploiting them was resource-intensive and costly. However, with the advent of AI, the expenses associated with vulnerability discovery, verification, and attack formulation have plummeted, rendering the previously effective, albeit cumbersome, cybersecurity safeguards obsolete. Presently, AI-powered vulnerability detection has emerged as a new battleground for top-tier AI firms, significantly speeding up the discovery process and uncovering a plethora of unexploited flaws in legacy code. Nevertheless, the speed at which vulnerabilities can be patched lags considerably behind their discovery rates, fundamentally transforming the dynamics of the cybersecurity attack-defense landscape. Looking ahead, widespread website shutdowns prompted by AI-discovered vulnerabilities may become a common occurrence.