Cisco Extends Zero Trust to AI Agents, Introduces DefenseClaw Framework
Cisco unveiled security innovations for AI agents, extending Zero Trust Access with agent discovery, Duo IAM identity management and model context policy enforcement in Secure Access SSE. It launched AI Defense: Explorer Edition for self-serve red teaming and introduced DefenseClaw, an open-source framework integrating NVIDIA OpenShell.
1. Zero Trust Access for AI Agents
Cisco has extended its Zero Trust Access controls to AI agents by introducing agent registration and discovery in Cisco Identity Intelligence, mapping agents to human owners in Duo IAM, and enforcing time-bound, fine-grained permissions through a model context protocol (MCP) gateway in Secure Access SSE.
2. AI Defense: Explorer Edition
The new self-service AI Defense: Explorer Edition leverages the core AI Defense Validation engine to allow developers, AppSec teams and security researchers to red team models and applications, measure risk posture and embed guardrails before deploying AI agents into production environments.
3. DefenseClaw Open-Source Framework
Cisco introduced DefenseClaw, an open-source secure agent framework that automates agent security and inventory. Plans include integration with NVIDIA OpenShell as a sandbox to streamline secure agent deployment and eliminate manual configuration steps.
4. Splunk AI Enhancements
New Splunk AI innovations within Cisco Secure Access automate security operations workflows, enabling SOC teams to detect and respond to AI-driven threats at machine speed, reducing manual intervention and accelerating incident response.