HCLTech and Palo Alto Networks: Securing Agentic AI by Design

Overview

Agentic AI goes beyond traditional chatbots by enabling autonomous Agents to plan and execute tasks across complex workflows using tools, APIs and contextual data. With the rise of platforms like Google Agentspace and A2A protocol, enterprises are increasingly adopting frameworks such as Agent Development Kit, LangChain and Semantic Kernel.

To help secure the evolving Agentic ecosystem, HCLTech and Palo Alto Networks have partnered to launch an Agentic solution that leverages Prisma® AIRS, a secure AI runtime platform designed to help enterprises confidently scale Agentic AI, with built-in security from the ground up.

The risks of Agentic AI without security

Expanding attack surface through APIs, tools and external systems

Supply-chain risks via open-source dependencies

Over-permissioned Agents making unauthorized decisions

Threats from Agent-to-Agent communications

Introducing Agents secured by Design powered by Prisma® AIRS - AI runtime security platform

HCLTech has developed an Agentic platform on Google Cloud that leverages Prisma® AIRS platform provided by Palo Alto Networks to provide comprehensive runtime security, threat detection mechanisms and Agent security posture management services.

The Prisma® AIRS platform seamlessly integrates into Agent workflows to deliver defense-in-depth security across:

Model scanning and software bill of materials (SBOM)
Agent security posture management
AI red teaming and threat simulation
Runtime LLM and Agent firewalls for prompt interception and tool call monitoring
Governance for A2A (Agent-to-Agent) and MCP (Model Context Protocol) interactions

Core capabilities

Assess and scan AI models

Ensure models from open source and managed services are safe and secure
Prevent malware from entering your environments
Stop execution of malicious code stored in the AI model
Secure model against model tampering, malicious scripts and deserialization attacks

Manage app and Agent posture

Continuously monitor and remediate your security posture
Prevent excessive permissions, sensitive data exposure, platform and access misconfigurations
Ensure a secure and compliant AI Agent and application user

Assess AI threats proactively with Red teaming

Holistic testing for predefined attack types
Detailed reporting on attacks, elaborating on sensitive information extraction
Real-time threat prevention via recommendations

Runtime security and AI Agent security

At network level via firewall
At code level via APIs and SDK
Stopping AI Agents specific attacks such as memory manipulation, tool misuse, hallucination attacks and more.

What we offer

Agent experimentation

Deploy Agents in sandboxed environments.
Model scanning and passive monitoring help identify risks without affecting performance.

Agent pilot build

Build scoped use cases (e.g., finance bots) with guardrails.
Apply red teaming, SBOM gates and privilege scoping to enforce best practices.

Launch production-grade Agents

Launch Agents into production with Agent governance and guardrails for transparency of Agent performance
Apply red teaming, SBOM gates and privilege scoping to enforce best practices.

Operate and optimize

Monitor Agent performance for drift (against specific KPIs)
Run cost optimizations
Maintain resilience with continuous red-teaming and threat telemetry feeding into HCLTech’s XDR/SIEM for 24×7 protection.

Unique offer

We offer a free Agentic AI Security workshop led by our COE. This can be followed by securing 2-3 simple Agents using an AI runtime security platform and can be used to demonstrate real-time threat detection proof-of-concept.