Zscaler CEO Jay Chaudhry on AI agent security, memory costs and AI infrastructure

Key Concepts

• Agentic AI: AI systems capable of performing tasks, making decisions, and interacting with applications autonomously.
• Zero Trust Security: A security framework based on the principle of "never trust, always verify," where no user or agent is trusted by default, regardless of their location.
• Exchange Technology: Zscaler’s cloud-based architecture that acts as a secure intermediary, validating connections between users/agents and applications.
• Infrastructure-as-a-Service (Security): A model where security is delivered via a global cloud infrastructure rather than physical hardware (firewalls/VPNs).

---

1. Main Topics and Key Points

• Securing Agentic AI: Zscaler is addressing the security risks associated with the rise of Agentic AI. As enterprises adopt AI agents, CIOs and CISOs are concerned about data loss and unauthorized access. Zscaler’s new product suite applies Zero Trust principles to these agents, ensuring they only access authorized applications.
• Infrastructure Resilience: Despite the global "memory chip crunch" and rising costs for processing power, Zscaler maintains that its impact is manageable. Unlike hardware-centric companies that ship physical firewalls, Zscaler operates a cloud-based infrastructure across 160 global locations, allowing for more efficient scaling.
• The Shift from Traditional Security: Jay Chaudhry argues that 30-year-old security models (firewalls and VPNs) are obsolete. These models rely on a "moat" mentality (keeping bad guys out), which fails in a modern environment where users and agents are distributed everywhere.

2. Real-World Applications and Market Context

• Enterprise Adoption: Zscaler currently serves over 45% of Fortune 500 companies, with over 50 million users utilizing their Zero Trust architecture.
• Market Disruption: The interview draws a parallel between the current AI revolution and the dot-com era. Just as the internet upended business models 25 years ago, Agentic AI is forcing a shift in how enterprises secure their digital environments.
• Model Creators as Tailwinds: The growth of companies like OpenAI and Anthropic is viewed as a positive "tailwind" for Zscaler. As these LLMs (Large Language Models) mature, enterprise adoption of agentic applications increases, which in turn drives the demand for specialized cybersecurity.

3. Methodologies and Frameworks

• The Zscaler Exchange Model:
  1. Traffic Interception: All traffic from users or agents is routed through Zscaler’s global exchange points.
  2. Validation: The system validates the identity and authorization of the entity (user or agent).
  3. Secure Connection: The entity is connected only to the specific application or service they are authorized to access, preventing lateral movement within the network.
• Scaling for Machine Speed: Because AI agents operate at "machine speed" without needing sleep or breaks, Zscaler’s architecture is designed to handle significantly higher traffic volumes than human-centric security models.

4. Key Arguments and Perspectives

• Agents as the New "Weakest Link": Chaudhry notes that while humans have traditionally been the weakest link in security, AI agents will soon take that role due to their speed and autonomy.
• Hardware vs. Cloud: Zscaler differentiates itself from "old school" cybersecurity firms by avoiding the physical hardware model. By operating in the cloud, they avoid the supply chain and cost volatility associated with shipping physical firewall boxes.
• Zero Trust as the Standard: The core argument is that in an era where "everyone is everywhere," the only viable security model is one that treats all entities as untrusted until verified.

5. Notable Quotes

• "Every one is excited about embracing AI, but all CIOs and CISOs are worried about how can they embrace it securely without causing issues such as data loss." — Jay Chaudhry
• "A user is the weakest link. Tomorrow, agents will be the weakest link. Agents run at machine speed." — Jay Chaudhry
• "Traditional security has been done using firewalls and VPNs... it's a very old model with 30-year-old architecture." — Jay Chaudhry

6. Synthesis and Conclusion

The transition to Agentic AI represents a fundamental shift in enterprise operations. Zscaler is positioning itself as the essential security layer for this transition by extending its proven Zero Trust architecture to autonomous agents. By moving away from legacy hardware-based security and focusing on a scalable, cloud-native exchange, Zscaler aims to mitigate the risks of data loss and unauthorized access in an increasingly automated digital landscape. The company views the maturation of AI model creators not as a threat, but as a catalyst for increased demand for their security services.