Palo Alto Networks Officially Acquires Portkey

Palo Alto Networks announced today that it has officially completed the acquisition of Portkey, a leader in AI gateway technology. The integration brings Portkey’s capabilities into Prisma AIRS as the new Prisma AIRS AI Gateway — a unified control plane built specifically to secure and govern AI agents operating at scale.

Enterprises are rapidly moving beyond simple copilots into deployments of autonomous agents that can reason, use tools, access sensitive data, and execute actions across systems. According to recent data cited in the announcement, 81% of enterprises are now either piloting or fully implementing AI agent solutions. These agents operate through APIs and MCP servers, creating new attack surfaces that traditional security tools struggle to address.

“The acquisition of Portkey accelerates our ability to deliver comprehensive AI security and governance at scale. We’re giving customers the visibility and control they need to confidently deploy agents across their environments.”

— Nikesh Arora

Chairman and CEO, Palo Alto Networks

The core challenge is visibility and accountability. Without a centralized layer, agent deployments often happen in silos, leading to inconsistent policies, unauthorized data access, and heightened security risks. Palo Alto Networks positions the new AI Gateway as the solution — a single enforcement point that identifies, authenticates, and authorizes every agentic interaction in real time.

Key features of the Prisma AIRS AI Gateway include a unified API for access to thousands of LLMs and agents, an agent registry, semantic routing, caching, artifact scanning, automated red teaming, and runtime behavioral monitoring. It also strengthens agent identity security through integration with Idira (formerly CyberArk), applying least-privilege controls to autonomous actions.

This acquisition reflects the growing industry focus on moving from fragmented point solutions to comprehensive platforms capable of handling the full agentic AI lifecycle — from development through production deployment and ongoing operation.

Key Terms

• Prisma AIRS AI Gateway: The new unified control plane resulting from the Portkey acquisition, providing centralized authentication, authorization, semantic routing, and runtime security for AI agents and LLM interactions.

• Agent Registry: A centralized catalog that discovers, inventories, and tracks all AI agents operating within the organization for visibility and policy enforcement.

• Unified AI Gateway: A single enforcement layer that intercepts agentic traffic, applies consistent security and governance policies, and supports thousands of LLMs and tools through a common API.

• Runtime Behavioral Monitoring: Continuous observation of agent actions during operation to detect anomalies, enforce guardrails, and prevent unauthorized data access or harmful behaviors.

• Idira Identity Security: Palo Alto Networks’ (formerly CyberArk) solution for governing non-human identities, now integrated to apply least-privilege controls to AI agents.

Conditions Driving This Change

• Enterprise adoption of AI agents has accelerated dramatically, with 81% of organizations now piloting or deploying them in production environments.

• Traditional security tools designed for static models and human users cannot effectively monitor or control autonomous agents that dynamically select tools and execute actions.

• Agent deployments are happening in fragmented ways across teams, creating inconsistent governance, visibility gaps, and increased risk of data exfiltration or policy violations.

• The proliferation of MCP servers and direct LLM API calls has expanded the attack surface, making centralized policy enforcement essential for both security and compliance.

• Regulatory pressure continues to mount, requiring organizations to demonstrate clear oversight, auditability, and accountability for high-impact AI systems under frameworks like the EU AI Act.

• Organizations face growing pressure from boards and CISOs to reduce shadow AI agent usage while maintaining innovation velocity.

• Legacy point solutions for LLM security and governance create operational complexity and do not scale to the volume and speed of agentic workflows.

• Security teams need solutions that combine discovery, real-time authorization, behavioral guardrails, and identity controls in one platform to match the pace of agent development.

"AI is evolving so rapidly that organizations often feel forced to choose between two failing strategies: scrambling to integrate a patchwork of 'point products' to stay current, or falling behind while waiting for legacy platforms to catch up. We're breaking that cycle. Palo Alto Networks is delivering a platform that stays on the cutting edge through a deliberate combination of organic innovation and strategic acquisitions. By making Portkey a critical component of the comprehensive Prisma AIRS platform, we do the heavy lifting of integration so our customers don't have to, enabling them to adopt the latest AI capabilities with speed and security."

Lee Klarich

Chief Product & Technology Officer

Palo Alto Networks

What AI Security Looked Like Before

Before the integration of Portkey’s technology into Prisma AIRS, securing AI agents was largely a fragmented and manual effort. Security and governance teams relied on a patchwork of point solutions — separate LLM gateways, basic API firewalls, custom scripts, and siloed identity tools. Many organizations had no centralized visibility into which agents were running, what tools they were using, or what data they were accessing.

Agent deployments often happened outside formal IT and security review processes. Teams would spin up agents using direct LLM API calls or standalone MCP servers, leading to shadow agent sprawl. Runtime monitoring was inconsistent at best, with limited ability to enforce least-privilege controls on autonomous actions. When issues arose — such as prompt injections, unauthorized data access, or unexpected tool chaining — detection and response depended heavily on after-the-fact log reviews rather than proactive, real-time enforcement.

This approach created significant operational friction. Developers faced delays waiting for security approvals, while security teams struggled to keep pace with the volume and velocity of agent development. The result was a dangerous gap between innovation speed and actual control.

"We joined Palo Alto Networks to bridge the trust gap that prevents AI from reaching its full potential. Our mission is to help enterprises move fearlessly. By combining our gateway with Palo Alto Networks AI security platform, we are helping organizations scale from experimental pilots to core business operations without compromising on safety or reliability."

Rohit Agarwal

CEO and Co-Founder

Portkey

What It Looks Like Now

With the completed acquisition of Portkey and its integration as the Prisma AIRS AI Gateway, enterprises now have a unified control plane for agentic AI. The gateway acts as a centralized enforcement layer that discovers, authenticates, authorizes, and monitors every agent interaction in real time.

Organizations can maintain a comprehensive agent registry, apply consistent policies across thousands of LLMs and tools, and benefit from semantic routing, caching, artifact scanning, automated red teaming, and behavioral monitoring. Integration with Idira (formerly CyberArk) brings strong identity and least-privilege controls specifically tailored to non-human AI agents.

This shift moves security from reactive, fragmented efforts to proactive, platform-level governance. Agents can now operate with appropriate autonomy while still remaining within defined guardrails. Security teams gain clear visibility and auditability without slowing down development teams.

The new Prisma AIRS AI Gateway provides the mission-critical control plane needed to scale autonomous workloads confidently, addressing the core challenges of visibility, policy enforcement, and runtime protection in one cohesive solution.