Palo Alto Networks Launches Prisma AIRS
As enterprise AI adoption accelerates across the board—from customer service to software development—so do the risks. In response, Palo Alto Networks has unveiled Prisma AIRS™, a comprehensive AI security platform aimed at protecting everything from large language models to AI-powered apps, agents, and sensitive data.
The company described Prisma AIRS as the most expansive solution of its kind, calling it the foundation for secure AI deployment in today's rapidly evolving threat landscape. Building on its Secure AI by Design initiative introduced last year, Prisma AIRS takes things to the next level by securing every layer of the AI stack.
In today’s enterprise environment, AI is everywhere—but so are the vulnerabilities. From prompt injections and hallucinated content to identity spoofing and API abuse, the list of AI-specific threats is growing fast. Palo Alto Networks says that with Prisma AIRS, organizations don’t have to choose between speed and security. The platform is designed to let teams build and deploy AI applications with confidence, while maintaining full visibility and control.
A Full Spectrum of Capabilities
Prisma AIRS isn’t just about locking down one aspect of AI. It's meant to cover the entire lifecycle, with key features including:
- AI Model Scanning: Identifies vulnerabilities in AI models before they’re put to use, guarding against tampering, embedded malware, and other attack vectors.
- Posture Management: Offers detailed visibility into configuration issues, over-permissive access, and exposure of sensitive data across the AI environment.
- AI Red Teaming: Simulates adversarial attacks using automated penetration tests that mimic how real-world attackers might exploit AI apps or models.
- Runtime Security: Defends deployed models and applications against active threats like prompt injection, hallucinations, malicious payloads, and data leakage.
- AI Agent Security: Addresses emerging risks tied to autonomous AI agents, especially those built on low-code/no-code tools, such as identity impersonation or misuse of internal tools.
Securing AI at the Speed of Innovation
Lee Klarich, the company’s Chief Product Officer, emphasized that AI is reshaping how people work and live, and said the attack surface is changing just as dramatically. Rather than rely on fragmented solutions, he said the answer lies in what he called platformization—building out a cohesive security architecture that can handle both traditional threats and the new wave of AI-specific vulnerabilities. Prisma AIRS, he noted, embodies that approach.
Anand Oswal, who leads the company’s network security division, echoed the urgency. He pointed out that many security teams still lack real-time insight into how their AI tools are being used—or potentially misused. Without that awareness, he warned, companies risk losing control over how decisions are made and data is handled. With Prisma AIRS, teams will be able to detect issues like attempts to extract sensitive data from LLMs or whether poisoned inputs are skewing model outputs.
Looking Ahead
The platform will be further strengthened by Palo Alto Networks’ recently announced plan to acquire Protect AI, a specialist in securing AI pipelines. That deal, pending standard closing conditions, is expected to wrap up in the first quarter of the company’s fiscal 2026.
For those curious about where Palo Alto Networks is heading next with AI security, the company is hosting a “Hello Tomorrow” livestream event on Tuesday, April 29, 2025, at 2:30 PM PT.
