Agentic Boom Exposes Gap in AI Security and Governance
The rapid adoption of agentic artificial intelligence is outpacing the deployment of generative AI models, indicating that organizations are seeking not just tools for content generation but agents that take action. This shift is primarily driven by the desire for enhanced productivity and revenue generation within enterprise environments.
The Acceleration of Agent Adoption
According to Daryan “D” Dehghanpisheh, the director of AI security at Palo Alto Networks, the adoption of agents is accelerating due to their direct impact on productivity and revenue. This rapid growth, however, is leading to a significant gap in governance and security readiness.
Challenges in AI Security Governance
One of the primary challenges identified is the lack of a comprehensive AI security governance framework. Dehghanpisheh emphasizes that as vendors introduce AI tools, they are increasingly scrutinized regarding their performance, security, safety, and governance. He notes, “Security, safety and governance are three distinctly different things. They’re related, but they are mutually exclusive.”
Key Discussion Points
In a recent video interview with Information Security Media Group, Dehghanpisheh elaborated on several important topics:
- Securing agent actions through least-privilege controls.
- Model security, testing, and insider risk.
- The emergence of universal AI vulnerabilities and gaps in coordinated disclosure.
Expertise in AI Security
Dehghanpisheh brings over 20 years of experience in technology and strategy to his role. His background spans various industries, including telecom and finance. He is recognized for his skills in management, leadership, strategy, and technology. Currently, he leads efforts in AI security at Palo Alto Networks, focusing on assembling teams of experts in AI, machine learning, business development, and digital marketing.
As organizations continue to embrace agentic AI, the need for robust governance and security measures has never been more critical. Addressing these gaps will be essential for safe and effective AI deployment in enterprise contexts.
