While artificial intelligence technology accelerates corporate management innovation, it also drastically expands the attack surface that hackers can target. Consequently, security experts have concluded that the traditional method of human detection and response to threats is no longer sufficient to defend against modern complex cyber attacks. Therefore, the industry is now pushing to build defense systems in the form of agents that can independently make judgments and act using AI; this signifies a fundamental shift in the security paradigm beyond mere technological adoption.
This trend was clearly demonstrated at the Next event hosted by Google Cloud. Morgan Adams of PwC emphasized that even if the attacker holds the advantage, the defense strategy can still succeed from the defender's perspective. He argued for integrating AI from the very beginning across all stages of security to maximize efficiency and identifying attackers as quickly as possible. Charles Kamakal, CTO of Mandiant, who also attended the event, diagnosed that the core challenge of corporate security has evolved into how quickly and safely AI can be integrated into defense systems. Mandiant, as a security company under Google Cloud, shares this same strategic direction.
A new threat that has emerged as particularly concerning in the industry is "Shadow AI." This refers to the phenomenon where employees independently create or deploy AI agents outside of official company approval or control systems. While companies encourage on-site experimentation to enhance productivity and innovation, the spread of AI without proper governance can lead to unforeseen security vulnerabilities. Morgan Adams emphasized that policies should not be redesigned as barriers that hinder innovation but rather as devices that help move faster and more safely. He explained that policies should not be designed to slow down speed, but to maintain necessary security barriers while encouraging innovation.
The problem lies in the fact that the time gap between attack and response is shrinking rapidly. According to Mandiant's latest report, attackers are compressing the response time available to defense organizations from hours to as little as 22 seconds. This means that relying on humans to understand the situation, go through reporting systems, and respond after a vulnerability is discovered is practically unsustainable. Charles Kamakal warned that while AI can enable responses on a scale far exceeding human capabilities, the risks inherent in AI itself always follow. Ultimately, the competitive landscape is becoming clearer: as attackers use AI to quickly find and exploit vulnerabilities, defenders must also use AI to identify anomalous behavior and accelerate patching.