Palo Alto Networks CEO Nikesh Arora has raised concerns about the cybersecurity risks posed by agentic artificial intelligence—AI systems that can act independently in the physical world. In an interview with CNBC’s Jim Cramer, Arora explained that while AI has already transformed industries through generative models, the next phase—where AI gains real-world autonomy—introduces unprecedented security challenges.

AI Moving Beyond Software: A New Era of Autonomous Agents

Arora pointed to Waymo’s self-driving cars as a prime example of agentic AI in action. Unlike traditional AI models that generate text or images, these AI-driven vehicles make real-time decisions about navigation, speed, and safety—essentially functioning as independent agents in the physical world.

“Giving AI arms and legs is the next frontier,” Arora noted. “When you step into a Waymo car, you're trusting AI to decide when to stop, when to turn, and how fast to drive. That’s a massive leap from what we’ve seen with generative AI.”

This shift from passive AI models to fully autonomous decision-makers is what makes agentic AI both powerful and potentially dangerous.

Cybersecurity Risks Grow as AI Autonomy Expands

With AI systems controlling critical infrastructure, from self-driving cars to automated factories and even financial systems, hackers now have more entry points than ever. Arora stressed that cybercriminals can exploit the same AI-powered tools used for efficiency and automation to launch sophisticated attacks.

At the Morgan Stanley Tech Conference in March, Arora highlighted this growing concern, stating:

“The development of agentic AI has opened a whole new threat vector. The biggest example? AI driving a car. Imagine the consequences if a hacker gains control.”

According to cybersecurity experts, the rapid evolution of AI systems means that many companies are unprepared for the security challenges they introduce. While organizations focus on harnessing AI for innovation, security protocols often lag behind—leaving massive vulnerabilities in AI-driven industries.

The Urgent Need for AI Security Measures

Arora emphasized the importance of securing AI-driven systems before they become mainstream. One of his key concerns is the reduction of human oversight—as AI takes over more tasks, the number of human access points for intervention shrinks. This makes it critical to ensure that AI agents can be trusted to operate without constant monitoring.

Government Collaboration & Automation in Security

Palo Alto Networks has already been working with government agencies to address these challenges. Arora revealed that his company has engaged with various parts of the U.S. government to develop strategies for securing AI-powered infrastructure.

He also sees automation as a key solution for improving efficiency and cutting costs. Citing the Department of Government Efficiency (DOGE), Arora explained that agencies need advanced automation to streamline operations and manage budgets more effectively.

“One of the biggest opportunities in cost-cutting is AI-driven automation,” he said. “The only way DOGE—or any government agency—can operate more efficiently is by automating a significant portion of their work.”

‍