Cisco extends SSE to agentic AI

Cisco said on June 3 that its Secure Access product is being extended to govern agentic AI workflows across large language models, Model Context Protocol servers, SaaS APIs and web destinations. In a Cisco blog post, the company said the move adds policy enforcement deeper into AI-driven task flows rather than limiting controls to the initial access decision. The announcement builds on Cisco’s March security push for what it calls the “agentic workforce,” when it tied Duo, Identity Intelligence, Secure Access and AI Defense more closely together. ### What exactly did Cisco say it added? Cisco’s June 3 post said Secure Access is evolving with “Agent Gateway,” which the company described as new functionality that extends policy enforcement across agent interaction with LLMs, MCP servers, SaaS APIs and web destinations. The post said organizations need to evaluate agent activity “throughout the workflow itself,” not only when access is first granted. (blogs.cisco.com) Cisco’s Secure Access product page describes the offering as a cloud-delivered security service edge, or SSE, platform grounded in zero trust. Cisco’s agentic AI security page says the company is tying identity context to access control and real-time behavior to address risks including overprivileged access and unpredictable actions by AI agents. ### Why does Cisco frame this as an SSE and identity problem? (blogs.cisco.com) Cisco’s own language centers on identity and traffic control rather than only model safety. The June 3 post said Secure Access is extending SSE and identity controls to help organizations govern agent actions across models, tools and web activity. Cisco’s March 23 announcement said Secure Access would add MCP policy enforcement and adaptive risk protection, while Duo and Identity Intelligence would handle agent discovery and agentic identity and access management. (cisco.com) Cisco’s broader SASE and Secure Access materials make the same case. A Cisco blog published in March said security in the “Agentic Era” requires protecting both agents from the world and the world from agents. A Cisco collateral page on the “agentic frontier” said organizations need visibility, control and performance across new agent interactions as traffic patterns shift across apps, APIs and enterprise systems. (blogs.cisco.com) ### What controls is Cisco saying enterprises need? Cisco’s June 3 post said the problem has shifted from access control to “action control.” The company said customers need ways to limit what agents can do after authentication, apply identity-aware checks during execution, and create audit trails for machine-driven tasks across tools and destinations. Cisco’s March launch materials described the same stack in more detail. (blogs.cisco.com) The company said it was establishing trusted identities for agents, enforcing Zero Trust access controls, hardening agents before deployment, and adding runtime guardrails so security teams could monitor and stop threats tied to autonomous actions. ### How does this fit with Cisco’s wider AI-security push? (blogs.cisco.com) Cisco used Cisco Live this week in Las Vegas to present a broader AI and infrastructure management strategy. Coverage from Network World and CRN said the company introduced new agentic AI, security and operations products as it tries to position itself around AI-driven enterprise infrastructure. (investor.cisco.com) Cisco also announced in March that it was expanding AI Defense and adding what it called AI-aware SASE capabilities. That release said the updates were designed to keep agentic workflows “safe, fast, and reliable,” linking AI governance to network and security controls rather than treating it as a standalone model issue. (networkworld.com) ### What should readers watch next? Cisco’s next public markers are likely to come through product documentation and event follow-ups tied to Secure Access, AI Defense and Cisco Live announcements. The company’s June 3 blog points customers to Agent Gateway inside Secure Access, while Cisco’s March 23 release and current product pages name Duo, Identity Intelligence and AI Defense as the adjacent pieces in the rollout. (blogs.cisco.com) (investor.cisco.com)