Article illustration 1

"Customers are concerned that if they don't have visibility to the agents, if they don't understand what credentials agents have, it's going to be the Wild West in their enterprise platforms."
— Nikesh Arora, CEO of Palo Alto Networks

As enterprises race to deploy AI agents—autonomous programs that access databases, invoke APIs, and execute tasks—a dangerous security gap is emerging. According to Palo Alto Networks CEO Nikesh Arora, traditional identity and access management (IAM) systems are catastrophically unprepared for this new wave of non-human workers that require privileged access to corporate systems.

The Privileged Access Blind Spot

Current security models focus primarily on privileged access management (PAM) for high-level human users, leaving approximately 90% of employee activities unmonitored due to cost constraints. This oversight becomes critical when AI agents—whether legitimate productivity tools or malicious implants—gain similar permissions:

  • AI agents performing retrieval-augmented generation (RAG) access sensitive databases
  • Agentic workflows using protocols like Model Context Protocol orchestrate multiple systems
  • Commercial software increasingly bundles autonomous functions requiring broad permissions

"An agent is also a privileged access user," Arora emphasized during a briefing. "Ideally, I want to know all of my non-human identities, and be able to find them in one place and trace them."

Expanding the Attack Surface

The identity crisis coincides with alarming threat escalation:
- 194,000+ malicious domains propagating AI-powered smishing (SMS phishing) attacks
- Automated credential theft targeting privileged accounts
- Nation-state actors weaponizing AI agents for infiltration

"We are seeing high-stakes credential attacks across the entire population of an enterprise," Arora warned. Current dashboards cannot track the explosion of machine identities, creating invisible pathways to critical assets.

The AI Vanguard Solution

Palo Alto's response involves two strategic pillars:

  1. Unified Identity Fabric: Integrating CyberArk's technology (acquired earlier this year) to create cohesive identity management spanning cloud, production workloads, and privileged access

  2. AI Guardians: Deploying Cortex AgentiX—AI security agents trained on 1.2 billion threat scenarios—to autonomously:

    • Hunt emerging attack patterns
    • Perform forensic analysis at scale
    • Provide SOC analysts with actionable intelligence

"You can't process terabytes of data manually," Arora noted. The system maintains human oversight initially, with progressive autonomy as trust develops: "I've done this five times with me watching it, it's doing it right, I'm going to allow it to act on my behalf."

The Paradox of Progress

This evolution represents a fundamental shift: The same AI capabilities creating unprecedented risks may become essential for defense. As enterprises confront their identity management debt, the organizations surviving the coming wave of AI-powered threats will be those that build security systems as dynamic and adaptive as the agents they aim to control.

Source: ZDNET interview with Nikesh Arora, November 2025