Zulfikar Ramzan, former CTO of RSA Security, highlights that agentic software introduces new risks by autonomously choosing paths to complete tasks. This capability complicates traditional security measures like least privilege, as agents often require broad access to various tools to function effectively.
During the Oracle event titled AI Cyberattacks Are Escalating: How to Secure Your Data Now, experts discussed the shrinking time window between initial system intrusion and a material data breach. Attackers are increasingly leveraging AI to accelerate the discovery and exploitation of vulnerabilities.
Ramzan emphasizes that organizations must prioritize data protection at the source as a final line of defense. Because human-speed triage cannot match machine-speed exploitation, security teams need to focus on strong authorization, continuous monitoring, and learning from past incidents.
To adapt, companies should map all active agents, identify their assigned identities, and audit the systems they can access. Integrating these insights into security programs allows teams to better prioritize defenses and manage the risks associated with multi-step automated tasks.
