Palo Alto Networks launches Prisma AIRS to secure enterprise AI

Palo Alto Networks has unveiled Prisma AIRS, an AI security platform designed to offer protection for the entire enterprise AI ecosystem, including applications, agents, models, and data.

The platform addresses the security needs of organisations adopting AI technologies at scale, particularly as enterprises increasingly employ AI applications and large language models (LLMs) in varied functions such as customer support and code generation.

Prisma AIRS provides several key security capabilities, including AI model scanning to detect vulnerabilities, posture management to monitor permissions and data exposures, AI red teaming to simulate attacks against AI systems, runtime security to guard against evolving threats during operation, and security for AI agents, including those built with no-code or low-code tools.

AI model scanning enables organisations to assess their AI models for vulnerabilities such as tampering, malicious scripts, and deserialization attacks. This component aims to help organisations adopt AI models safely by identifying security risks before deployment.

Posture management provides insights into security risks associated with an enterprise's AI ecosystem. It highlights issues such as excessive permissions, sensitive data exposure, and both platform and access misconfigurations.

The AI red teaming feature allows organisations to perform automated penetration testing on AI applications and models by utilising a red teaming agent. This agent stress-tests AI deployments, learning and adapting like a real attacker to uncover potential exposures and lurking risks before malicious actors can exploit them.

Runtime security within Prisma AIRS protects AI applications powered by LLMs against threats encountered during operation. These threats may include prompt injection, malicious code execution, toxic content, sensitive data leaks, resource overload, and hallucination.

To address risks associated with AI agents, Prisma AIRS includes safeguards for both standard and no-code/low-code platforms. It aims to defend against agentic threats such as identity impersonation, memory manipulation, and misuse of tools.

Lee Klarich, Chief Product Officer for Palo Alto Networks, said, "AI agents and apps are transforming the way we work and live. In parallel, the attack surface isn't just expanding, it's fundamentally changing. The last thing organizations need is more point products to secure their use of AI. Organizations need best-in-class security delivered via the right architecture - platformization is that architecture. Prisma AIRS addresses both traditional and AI specific threats with best-in-class security capabilities delivered in a comprehensive, unified AI security platform that enables organizations to deploy AI bravely."

Anand Oswal, Senior Vice President and General Manager at Palo Alto Networks, added, "As organizations integrate AI into every aspect of their operations, securing it requires a runtime security platform that provides continuous visibility and real-time insight. Without this, security teams are left in the dark about how AI is being used, misused, or manipulated, which puts critical data and decisions at risk. Prisma AIRS empowers teams with answers to essential questions, like whether someone is exploiting an LLM to extract sensitive information or if a compromised API is feeding the model poisoned data. These insights are vital to maintaining trust and safeguarding AI."

Palo Alto Networks indicated that Prisma AIRS would be enhanced by the company's plans to acquire Protect AI. The acquisition is subject to customary closing conditions and is expected to be finalised in the first quarter of the company's fiscal 2026. Protect AI focuses on securing the use of AI, aligning with Palo Alto Networks' efforts to address AI security concerns comprehensively.