Palo Alto Networks has introduced Prisma AIRS, a groundbreaking AI security platform designed to fortify the entire enterprise AI ecosystem. This new initiative aims to safeguard AI applications, agents, models, and the sensitive data they utilize at every juncture. As organizations increasingly integrate AI into their operations, the need for robust security measures has become critically apparent.
This launch builds on the company’s Secure AI by Design portfolio unveiled last year, positioning Prisma AIRS as a vital tool for businesses wishing to deploy AI with confidence. The platform addresses the growing vulnerabilities associated with rapid AI adoption, which, while driving innovation in various sectors—from customer support to code generation—also exposes businesses to new security risks.
As enterprises embrace AI technologies, they are confronted with a plethora of challenges, including potential security blind spots, risks, and vulnerabilities. With the increasing deployment of AI applications and large language models (LLMs) across numerous functions, there’s a heightened need for a comprehensive AI security platform that can effectively mitigate risks and prevent security incidents.
The capabilities of Prisma AIRS are extensive, targeting multiple areas of vulnerability within an organization’s AI infrastructure. Key features include:
-
AI Model Scanning: The platform allows organizations to safely adopt AI models by scanning them for vulnerabilities. This feature secures the AI ecosystem against risks such as model tampering, the injection of malicious scripts, and deserialization attacks, ensuring a fortified AI deployment.
-
Posture Management: Prisma AIRS enables businesses to gain valuable insights into security posture risks within their AI ecosystem. This includes the identification of excessive permissions, exposed sensitive data, and various types of misconfigurations, helping organizations address vulnerabilities proactively.
-
AI Red Teaming: The platform includes tools for uncovering potential exposure and risks before they can be exploited by malicious actors. Organizations can conduct automated penetration tests on AI applications and models. The Red Teaming agent simulates real-world attacks, stress-testing AI deployments and adapting strategies as a legitimate hacker would, thereby exposing vulnerabilities.
-
Runtime Security: Prisma AIRS offers runtime protection for LLM-powered AI applications, safeguarding them against a variety of threats. These threats include prompt injections, malicious code executions, toxic content, sensitive data leaks, resource overloads, and even hallucination, thereby ensuring system integrity.
- AI Agent Security: The platform extends its security measures to agents, including those developed on no-code or low-code platforms. This feature addresses emerging threats posed by identity impersonation, memory manipulation, and misuse of tools, thereby fortifying the agentic aspects of AI ecosystems.
Lee Klarich, Chief Product Officer at Palo Alto Networks, emphasized the transformative impact of AI agents and applications on workplace dynamics. He noted that as the attack surface evolves, organizations face the pressing need for a more integrated approach to security rather than relying on disparate point products. “Organizations need security delivered via the right architecture—platformization is that architecture. Prisma AIRS tackles both traditional and AI-specific threats through a unified platform, enabling organizations to confidently deploy AI,” he stated.
Additionally, Anand Oswal, Senior Vice President and General Manager at Palo Alto Networks, remarked on the necessity for a runtime security platform that ensures continuous visibility and real-time insights into AI usage. He pointed out that without such transparency, security teams may remain oblivious to how AI is being manipulated, putting crucial data and decision-making processes at risk. “Prisma AIRS equips teams with answers to critical questions about AI exploitation and data integrity, essential for maintaining trust and safeguarding sensitive information,” he concluded.
Palo Alto Networks has further underscored its commitment to advancing AI security by announcing its intent to acquire Protect AI, an innovative leader in safeguarding AI usage. This acquisition is pending customary closing conditions and is projected to conclude by the first quarter of fiscal 2026. The addition of Protect AI’s capabilities is expected to significantly enhance Prisma AIRS, providing organizations with even greater security measures as they navigate the complexities of AI integration.