Agentic AI Amplifies Cyberattacks: Experts Urge AI-Driven Defenses and Vendor Patching Mechanisms

The rapid advancement of agentic AI has significantly increased both the speed and breadth of cyberattacks. During the cybersecurity incident response symposium held on November 3, 2026, industry, government, and academic experts discussed emerging intelligent attack vectors. Ray Liu, associate director at Chunghwa CyberSecurity International, noted that corporate security has traditionally concentrated on network perimeters, but recent breach cases show that the trust boundary has shifted toward credentials and tools. Scholars noted instances where an automated AI framework attempted over 2,900 logins on Microsoft accounts within one minute, emphasizing the necessity of AI-speed defenses and tiered automated mitigation. Regarding defensive frameworks, Sun Wei-zhe, director at the National Institute of Cyber Security (NICS), stated that the institute is promoting the PSCIRT (Product Security Incident Response Team) framework to hold product vendors accountable for rapid vulnerability patching, with plans to establish institutional guidelines across Taiwanese enterprises.

Key takeaways

• Agentic AI increases attack frequency, with experts highlighting automated frameworks attempting over 2,900 logins per minute, necessitating AI-speed countermeasures.
• The security trust boundary has shifted from traditional network perimeters to identity credentials and access tools, requiring automated access control and blocking.
• Taiwan's NICS is promoting PSCIRT to ensure vendors take responsibility for timely patching and to develop formal security frameworks for enterprises.