What is Prisma AIRS? How Palo Alto Networks secures the AI runtime environment

As artificial intelligence transforms enterprise software, a new attack surface is emerging: the AI itself. From inference pipelines to model artifacts, the components powering generative AI and machine learning are increasingly becoming targets for malicious actors. In response, Palo Alto Networks (NASDAQ: PANW) has launched Prisma AIRS — a purpose-built AI Runtime Security solution designed to discover, monitor, and secure AI models across development and production environments.

With AI systems moving beyond passive data analysis into agentic autonomy, cybersecurity strategies must evolve. Prisma AIRS represents Palo Alto Networks’ strategic bet that defending the infrastructure around AI is not enough. The models, data, artifacts, and runtime interactions must also be protected — continuously, contextually, and at machine speed.

Why Prisma AIRS, and Why Now?

The rise of generative AI has dramatically accelerated enterprise interest in deploying large language models, vision models, and domain-specific AI tools across use cases. But these models often operate in environments not designed with security in mind. As noted by Palo Alto CEO Nikesh Arora during the Q3 FY2025 earnings call, even Palo Alto Networks itself is running over 35 different AI models across its own products.

This proliferation introduces risks ranging from exposed model weights and data poisoning to prompt injection and adversarial manipulation. Yet many organizations still lack visibility into which AI models they’re running, where the data is coming from, and how secure the deployment pipeline truly is. Prisma AIRS addresses this gap. It helps enterprises discover, scan, and protect AI models and their associated infrastructure. According to Arora, customer demand is already strong, with the company reporting an eight-figure pipeline for AIRS shortly after launch. This underscores the urgency with which enterprises are seeking solutions tailored to the AI era.

What Does Prisma AIRS Actually Do?

Prisma AIRS is a runtime security platform for artificial intelligence. It goes beyond static posture checks by embedding real-time security controls directly into AI workflows. The platform provides visibility into model behavior, data flow, and operational interactions whether the models are running in containers, virtual machines, or cloud-native environments.

It begins by automatically detecting and cataloging AI components across cloud and on-prem environments. These include model files, weights, training datasets, and configuration pipelines. Once discovered, the system performs advanced scanning to uncover vulnerabilities in AI logic or deployment, including validation routines, robustness checks, and exposure analysis. In live production, Prisma AIRS tracks interactions with models — such as API calls and user behavior — and flags anomalies. When suspicious behavior is detected, it initiates automated remediation steps. Enterprises can also define guardrails to protect training data and enforce compliance requirements such as encryption protocols, access governance, and data residency adherence.

How Is Prisma AIRS Positioned in the Broader Platform?

Prisma AIRS is tightly integrated into Palo Alto Networks’ Cortex XSIAM platform and Prisma Cloud offering. This integration allows organizations to apply the same detection, response, and automation workflows to AI environments that they already apply to their broader security stack. It ensures seamless adoption, consistent policy enforcement, and unified reporting.

AIRS plays a critical role in Palo Alto’s long-term platformization strategy. All telemetry — including that generated by AI models — flows into a central data lake where it can be analyzed and acted upon by the company’s machine learning engines. This consolidation removes operational silos and increases detection efficacy. According to Chief Product Officer Lee Klarich, AIRS is essential in enabling secure agentic AI systems, where AI not only interprets data but acts on it autonomously. With Palo Alto’s Prisma Access Browser now securing over three million endpoints, AIRS will increasingly operate in runtime environments powered by embedded AI agents — where security must be native to the application layer.

What’s the Role of Protect.ai in the Prisma AIRS Vision?

To accelerate the Prisma AIRS roadmap, Palo Alto announced the acquisition of Protect.ai, a startup specializing in model scanning and AI red teaming. The $700 million deal is expected to close by Q1 FY2026 and will bring deep expertise into Palo Alto’s AI security portfolio. Protect.ai’s capabilities will enhance AIRS with tools for adversarial testing, automated threat simulation, and auditing frameworks tailored to emerging global AI compliance standards.

Nikesh Arora confirmed that the Protect.ai team will take over product leadership for Palo Alto’s broader AI security line. This reflects a strategic commitment to AI security as a core pillar of the company’s future. The acquisition also gives Palo Alto credibility in a highly technical and rapidly evolving domain, where domain specialization is critical to customer trust.

What Are Analysts Saying About Prisma AIRS?

Though Prisma AIRS is in early commercialization, analysts covering Palo Alto Networks have highlighted it as a timely and forward-thinking initiative. In Q&A sessions following the Q3 FY2025 earnings call, analysts from Goldman Sachs and Morgan Stanley noted that Prisma AIRS differentiates Palo Alto from traditional security vendors by addressing a fast-growing need in AI infrastructure protection. They emphasized that Palo Alto is one of the first companies explicitly addressing the security of AI systems as agents — not just as data pipelines.

According to Klarich, AIRS allows security teams to apply threat models not just to the perimeter or the cloud, but to the behavior, reasoning, and permissions of AI models themselves. This shift toward behavioral and semantic AI security is a profound change in how organizations will be expected to think about cyber defense in the near future.

How Big Is the Market for AI Runtime Security?

While no analyst firm has defined a dedicated AI runtime security segment yet, the consensus is that this category will grow rapidly. The convergence of AI adoption, cloud-native infrastructure, and regulatory pressure is already creating customer demand for real-time observability, threat detection, and governance capabilities specific to AI. Palo Alto’s own AI-related ARR hit $400 million in Q3 FY2025 — a figure that could scale rapidly with the rollout of AIRS and the addition of Protect.ai technologies.

Palo Alto Networks is positioning itself as a first mover with broad platform integration, deep enterprise relationships, and early traction. This head start could allow it to define the AI runtime security market before competitors even fully articulate the threat model.

Should Enterprises Adopt Prisma AIRS?

For enterprises building or deploying AI systems, Prisma AIRS offers a comprehensive and operationally mature option for securing the runtime layer. Unlike traditional tools focused on static scanning or infrastructure posture, AIRS embeds observability and response directly into the application workflow. This makes it uniquely suited for environments where AI is dynamic, autonomous, and continuously interacting with users and systems.

Whether the enterprise is building internal copilots, deploying LLMs in customer-facing interfaces, or experimenting with multi-agent AI systems, the risks associated with those models are no longer theoretical. Prisma AIRS offers a practical path toward securing those environments, ensuring that innovation does not come at the cost of integrity, compliance, or control.

As AI becomes a first-class citizen in enterprise infrastructure, Prisma AIRS may prove to be one of the most important product launches in Palo Alto Networks’ history.