AI's Double-Edged Sword: Innovation, Risk, and the Expanding Attack Surface

""Agents are not just assistants," Khatri explained. "They're systems you give a task to and expect results - end to end.""

"Modern large language models now operate at the level of reasoning systems, capable of research, code generation, and structured problem-solving. According to Khatri, the industry is now entering the age of AI agents, which are systems that can execute tasks autonomously with minimal human input."

"This shift matters because autonomy changes the risk profile entirely. When AI systems can plan, iterate, and act on their own, they can just as easily optimize attacks as they can optimize workflows."

"Khatri highlighted how early generative models enabled highly personalized phishing and social engineering attacks. These weren't generic scam emails; instead, they were tailored, researched, and convincingly human. Today, generative video and audio models have taken this further, enabling deepfake calls a"

AI systems are reshaping cybersecurity, fraud, misinformation, and physical security as they become more capable and dangerous. Early scripted assistants have been replaced by large language models that can reason, research, generate code, and solve structured problems. The shift toward AI agents enables autonomous task execution with minimal human input, changing the risk profile because systems can plan, iterate, and act. Intelligence is becoming an attack surface, enabling highly personalized phishing and social engineering through generative models. Generative video and audio models further enable deepfake calls and more convincing impersonation, increasing the scale and effectiveness of attacks.

Read at Medium