Prague: As cyber threats become more refined and the workforce grows outside the perimeter of traditional offices, the older models of network security that were automatically trusted by everyone in the system are no longer sufficient. Traditionally, security worked like a “castle and moat.” Once inside the corporate network (castle), you were trusted and the moat (firewall and perimeter defense) kept outsiders at bay. However, Jay Chaudhry, India-born founder and CEO of cybersecurity firm Zscaler, explained that the model is failing in today's cloud-first world where users are no longer confined to physical locations.One widely accepted new approach that is changing cybersecurity globally is what is called zero trust security. This approach does not allow you to roam the castle freely, according to Chaudhry. “We only escort them to meeting rooms where they are permitted, and once they are done, we escort them. Moat's approach is no longer kept well,” he explained in his keynote address at Zenith Live 2025. He noted that as AI accelerates, the zero trust approach has become even more important. Because hackers can use AI to deceive systems. Zero Trust works on principles. Don't trust it, don't always verify it. Users, devices, connections, etc.“Data is everywhere. We serve over 50 million users worldwide. Our employees access tools such as Salesforce, Microsoft 365, SAP, LinkedIn, YouTube and more through secure platforms,” Chaudhry said. “Think of us as an international checkpoint. We record all transactions. Who accessed it, when and where? These digital records act like visitor logs and can capture early signs of malicious activity. It processes approximately 500 billion transactions every day and generates 500 trillion data points. This is half. Analysis is like finding a needle in a haystack. That's where generative AI comes in handy. We discover tools such as violation prediction that will halt the threat before it unfolds,” Chaudhry said.He said the same principles must extend to the following frontiers: Agent AI. “Agent AI can act autonomously and have access to a wide range of applications and data. As a result, the need for zero trust is more urgent than ever,” stressed Chaudhry. To address this challenge, he said Zscaler is working with several partners, including Microsoft, to define and manage the identity of AI agents. “When agents are established, they are expanding their platform to apply the Zero Trust Principle, so they not only protect users, but also AI agents acting on their behalf,” he said.Phil Tee is a pioneer in using AI in IT operations at venture Moogsoft (acquired by Dell in 2023), and joined Zscaler earlier this year as head of AI innovation. He said that if the company has a security incident, it could take months or months to diagnose what happened and take corrective action. “But if you can use AI to extract the answers from the logs, you can reduce it from months to minutes to zero. And it can be combined with an agent architecture and can be automated healing, perhaps an autonomous security system,” he said.Given the use of AI tools by many employees, the company is also working on solutions for deep, rapid inspections, just as cybersecurity tools do deep packet inspections to identify viruses and malware. This is to ensure that there is no accidental data removal that allows the engineer to paste some source code into the underlying model, for example, because the engineer is trying to debug or optimize. That data can find a way to the model's training set. “You can see the interactions between clients, like the copilot, for example. You can use the language model to see the prompts to see if it's just text.(The correspondent was in Prague at the invitation of Zscaler.)
