Palo Alto Networks launches Cortex XDR One for AI agents

Palo Alto Networks has launched Cortex XDR One, a platform that combines endpoint detection and response with protection for autonomous AI agents.

The product is intended to address risks emerging as businesses allow AI agents to interact with internal systems such as databases, production code, financial ledgers and customer management platforms.

Cortex XDR One brings together extended detection and response and Cortex Agentic Endpoint Security in a single interface and sensor. It is designed to cover threats aimed at both traditional operating systems and AI-driven tools used on enterprise endpoints.

The launch reflects a broader shift in corporate security priorities as AI coding agents, copilots, Model Context Protocol servers and personal AI tools become more common in day-to-day operations. Palo Alto Networks says these systems can inherit user permissions and act in non-deterministic ways, creating openings for prompt injection, indirect data poisoning and unauthorised API calls.

Traditional endpoint detection and response tools were not built to manage those risks. Basic software containers also do not fully address the problem when AI agents can access sensitive systems and data through legitimate credentials.

Platform design

The platform offers real-time visibility into AI software installed on endpoints, including unauthorised tools sometimes described as shadow AI. It also includes execution controls intended to apply checks at both the AI runtime layer and the operating system layer.

The product uses small language models on the sensor itself to inspect and block fileless threats even when a device is offline. By processing data locally, organisations can examine suspicious activity without moving sensitive code to the cloud.

Another element is a single sensor that gathers telemetry across endpoint activity. Palo Alto Networks says this allows security teams to connect attack chains across different domains, reduce false positives by up to 95%, and support automated remediation.

Customers can deploy the agentic endpoint security element as a standalone module alongside third-party endpoint detection and response tools, or use it through the full Cortex XDR One platform. The offer can also be paired with Unit 42 managed detection and response services, which now include managed agentic endpoint security.

AI security focus

The technology behind Cortex Agentic Endpoint Security comes from Palo Alto Networks' integration of Koi, a startup focused on agentic security. The tools were first used internally by the company's information security team before being incorporated into the broader platform.

Orcun Tezel, Vice President, Technical Solutions, Asia Pacific & Japan, Palo Alto Networks, said the rise of AI agents on enterprise devices requires a change in how organisations think about endpoint protection.

"The expansion of agentic capabilities across enterprise endpoints demands a security posture that treats AI agents with the same rigour and control as traditional system processes," Tezel said.

He said the company sees the product as an extension of earlier efforts to broaden detection and response beyond a single environment.

"Just as we pioneered XDR to protect network, identity, and cloud environments, Cortex XDR One expands endpoint protection to govern AI agents at the source, enforce OS-level boundaries, and stop evasive threats in real time," Tezel said.

Cortex XDR One and Cortex Agentic Endpoint Security are available immediately.