What happened
On Oct 15, Palo Alto Networks announced Prisma AIRS innovations: (1) runtime protection integrated directly into Claude Enterprise via inference hooks covering Claude Code, Claude.ai and Claude Cowork — applying DLP/PII policies and blocking prompt injection/jailbreaks before inference; (2) expansion of AI Supply Chain Security to scan agent skills, plugins, MCP servers and open-source deps (3M+ third-party artifacts pre-scanned) beyond models; (3) AI Gateway enhancements: bidirectional MCP Guardrails (blocks concealed instructions, redacts sensitive data in tool outputs) and Signals & Smart Routing for cost/quality optimization.
Why it matters
This is a major vendor extending AI security from the gateway to the model-inference layer of a frontier coding assistant (Claude Enterprise), plus expanding supply-chain coverage to agent skills/MCP — the fastest-growing agent attack surface. It directly competes with/integrates alongside the Anthropic ecosystem and positions Prisma AIRS as an end-to-end (dev→runtime) agent-security platform with 3M+ artifacts pre-scanned.
Applicability
Enterprises running Claude Enterprise/Claude Code for coding and workflow automation should evaluate Prisma AIRS integration for DLP extension and prompt-injection blocking; AI supply-chain teams adopting MCP servers should assess AIRS's MCP scanning and gateway guardrails.