Introducing Prisma AIRS AI Supply Chain Security

Nov 20, 2026
4 minutes

Secure Every Model and Agent Skill Before It Reaches Production

We are pleased to announce the general availability of Prisma®  AIRS™ AI Supply Chain Security. It secures every model and agent skill your AI is built on, from one platform, before any of it reaches production.

Modern AI applications are assembled, not built from scratch. Developers pull in open-source models and third-party agent skills that can run code, access data and use credentials. Each one becomes part of the AI supply chain, and its risks become the organization's.

The question is no longer whether teams will use third-party AI artifacts. It is how to verify them without slowing development down. AI Supply Chain Security scans models and agent skills inside the pipelines teams already use, stopping threats and policy violations while teams build.

Why AI Artifacts Need Their Own Security

Traditional application security tools are largely designed for the code developers write. They are not designed for the AI artifacts teams download and deploy. We aren't just talking about libraries and packages. We are talking about model files that can carry hidden code, and agent skills that can carry instructions an agent will follow.

Most organizations face three gaps when they try to address that risk:

  • Fragmented coverage. Many tools cover models or skills, not both, leaving security teams to piece together separate products.
  • Invisible threats. Code scanners typically miss backdoors in model files and malicious instructions embedded in skills.
  • Policy gaps. Models with license issues, artifacts from risky sources and unapproved skill versions can reach production without review.

As a result, models and skills can reach production uninspected, and a single compromised artifact can spread to every application and agent that uses it.

One Platform for Models and Agent Skills

Prisma AIRS AI Supply Chain Security builds on our AI Model Security capabilities and extends them to agent skills. Security teams get one place to validate every artifact before it is deployed.

  • Unified coverage. Secure models and skills from one interface in Strata Cloud Manager, with checks purpose-built for each artifact type. Security Groups define what "secure enough" means for each team, environment or use case.
  • AI-specific threat detection. For models, detect code that runs at load time or during inference, architectural backdoors and unsafe serialization formats across 35+ model formats. For skills, detect embedded instructions that try to override agent behavior, arbitrary code execution, credential access and data exfiltration paths.
  • Policy Enforcement. We let you know when to block an artifact that fails license, source or version rules before it reaches production. You can apply strict blocking in production and monitoring in development, so teams keep visibility without slowing experimentation.

AI Supply Chain Security works where developers already build. Model scans run locally through the Python SDK and CLI, and plug into CI/CD pipelines that provide gateways to deployment within organizations. Proprietary models never leave the customer's environment; only scan metadata and results are sent for policy evaluation. Skills can be scanned from public GitHub repositories, through the SDK or by ZIP upload.

Why Prisma AIRS

AI Supply Chain Security is backed by the depth of the Prisma AIRS platform and Palo Alto Networks threat research:

  • 3M+ models prescanned on Hugging Face. Prisma AIRS continuously indexes public models, so teams can check many of them against policy immediately.
  • Local model scanning. Sensitive models stay inside your environment.
  • Continuous threat intelligence. Detection is updated with research from Unit 42 as new AI attack techniques emerge.

Existing AI Model Security customers can add agent skill scanning under their current license at no extra cost, at launch. Skill scanning is currently available in North America regions and expanding to other regions in the near future.

AI Supply Chain Security is part of Prisma AIRS, the industry's most comprehensive AI security platform, alongside AI Red Teaming, AI Runtime Security, Agentic Endpoint Security, AI Posture Management, AI Gateway and Agent Identity Security. Together, they help enterprises discover, assess, protect and govern AI from pilot to production.

AI is increasingly sourced from components built by third-parties but modified by your teams. With AI Supply Chain Security, every one of them can be verified before it runs.

Learn more about Prisma AIRS AI Supply Chain Security or read our solutions brief.