Cisco Ships Zero Trust for AI Agents, Self-Service Red Teaming, and Agentic SOC Tools at RSAC 2026

Cisco unveiled a suite of enterprise security products at RSAC 2026 focused on securing AI agent workloads. Key announcements include Zero Trust Access for AI agents with new Duo IAM capabilities for registering agents and enforcing task-specific permissions via an MCP gateway, and AI Defense: Explorer Edition — a self-service adversarial testing tool supporting prompt injection, jailbreak testing, and CI/CD integration with GitHub Actions, GitLab, and Jenkins. An Agent Runtime SDK embeds policy enforcement at build time across AWS Bedrock, Google Vertex, Azure AI Foundry, and LangChain. Cisco also released DefenseClaw, an open-source agent security framework integrating with NVIDIA's OpenShell. On the SOC side, Splunk is shipping Detection Studio, Federated Search, Exposure Analytics, and several specialized AI agents including a Triage Agent and Malware Threat Reversing Agent, with general availability rolling out through June 2026.