Know what agentic AI can access, Zscaler warns

Cloud-native security provider Zscaler has joined the list of sponsors for the annual ITWeb Security Summit 2026 in Johannesburg, where it will deliver a presentation on enabling agentic AI-data security.

Zscaler operates the AI-powered Zero Trust Exchange to protect users and applications. Its core business is security service edge, replacing traditional firewalls and VPNs with identity-based access.

The company says organisations are adopting AI agents at breakneck speed. But for enterprise AI agents to be useful, they need access to business data – and that data may be at risk without appropriate guardrails.

Organisations must understand what data agents can access, what they can do with it (read/write) and whether compliance requirements are satisfied.

Srikanth Nellore, senior director of product management at Zscaler, is scheduled to present at the summit. He will explain how the company can identify data exposed to AI agents across platforms such as AWS Bedrock, Microsoft Foundry and GCP Vertex, and enforce security policies at runtime.

The company believes the topic is uniquely relevant because it addresses the biggest shift in enterprise technology this year: the move from chat-based AI (GenAI) to autonomous AI (agentic AI) and the potential security risk to corporate data.

“Agentic AI should be implemented with guardrails from the start, not as an afterthought,” Nellore explains. “Organisations need to adapt their zero-trust principles to ensure AI agents and APIs don’t have unfettered access to internal systems. Every API or AI agent should be treated as an identity and assigned to a group with limited access.”

Zscaler says the world is becoming increasingly risky, with AI-driven cyber attacks, quantum computing threats, geopolitical tensions and supply chain volatility impacting businesses.

For organisations aiming to maintain continuity and agility, reacting is no longer sufficient – resilience must shift from internal defence to an outward-facing design principle, the company says.

Citing its own research, Zscaler notes that legacy architecture remains a major challenge: 81% of organisations still rely on legacy systems; 64% say infrastructure limits failure response; and 59% report it cannot keep pace with business change.

To build resilience, organisations must conduct external stress testing – simulating quantum disruption, AI innovation and supplier interdependence – to identify hidden risks.

See also

“AI-driven attacks operating at machine speed require AI-native defences,” Nellore adds. “Fragile supply chains demand a zero-trust approach to prevent lateral movement from compromised partners.

“The global skills gap is being addressed through platform consolidation and automation, enabling lean security teams. This shifts resilience from building bigger walls to creating agile, invisible architectures based on identity and automation.”