With comprehensive security solutions covering the entire lifecycle of AI agents, Cisco helps organizations confidently deploy AI on a large scale.
Recently, at the RSA Conference 2026, Cisco introduced solutions to address AI security challenges and remove one of the biggest barriers to deploying AI agents.
By establishing trusted identities, implementing strict access control based on the Zero Trust model, enhancing security for agents before deployment, and enforcing control mechanisms during operation, Cisco is progressively integrating security from the very foundation of its AI ecosystem. Simultaneously, the company is equipping its Security Operations Centers (SOCs) with the necessary tools to counter threats at machine speed.
"AI agents are not only making current work faster; they are forming a new workforce – 'digital colleagues' who have the potential to significantly expand what organizations can achieve," said Jeetu Patel, President and Chief Product Officer of Cisco.
"Projects that were previously delayed due to a lack of resources are now within reach. The only limit is imagination, and security teams are key to unlocking this opportunity by ensuring the agent workforce is secure enough to be trusted," Jeetu Patel emphasized.
In a recent Cisco survey of large enterprise customers, 85% said they were experimenting with AI agents, but only 5% were actually implementing the technology in practice.
Protecting the world from AI threats: Building trust before the threats begin to operate.
According to the recently published Cisco Talos Year in Review 2025 report, attackers primarily target a group of critical components, including user authentication systems, access control decision enforcement, or those acting as intermediaries to establish trust between systems. This identity-focused trend is projected to continue to increase with the growth of AI-powered workloads.
To address these challenges, Cisco has extended its Zero Trust model to AI agents, ensuring that these agents are tied to the responsibilities of a specific personnel and that their actions are closely monitored.
Duo's new Identity and Access Management (IAM) capabilities are integrated with Model-Contextual Policy Enforcement (MCP) and behavior-based monitoring capabilities within Cisco Secure Access, enabling the application of rigorous access controls. Through this, Cisco helps organizations gain, for the first time, comprehensive visibility and effective governance of their AI workforce.
Protecting AI agents from external impacts: AI Defense safeguards the agent workforce.
To help more organizations proactively address this challenge, Cisco is popularizing the industry-leading capabilities of its AI Defense solution with the launch of Cisco AI Defense: Explorer Edition. This new self-service solution is built on the same core AI Defense Validation platform that is trusted by Global 2000 companies.
After registration, users can begin conducting red teaming (attack testing) on AI models and applications intended for deployment in agent processes, to detect vulnerabilities and assess risk levels before deployment. This toolkit enables AI developers, application security teams (AppSec), and security researchers to build and protect AI agents effectively.
Cisco has now launched DefenseClaw—a platform for security agents designed to eliminate the barrier between development and security. By integrating a suite of essential open-source tools, including Skills Scanner, MCP Scanner, AI BoM, and CodeGuard, DefenseClaw enables developers to deploy security agents with greater speed and reliability.
DefenseClaw's features will integrate directly with NVIDIA's OpenShell, expanding the existing partnership to deliver robust and automated security capabilities at the runtime level.
Detecting and responding to machine speed: Enhancing SOC capabilities towards agent-based operation (agent SOC)
AI technologies are a double-edged sword. As the latest Talos Year in Review report shows, vulnerabilities like React2Shell were exploited almost immediately and automatically, most likely driven by the use of agent AI to build new exploit toolkits.
Splunk, part of Cisco's security solutions portfolio, has been integrating AI capabilities into critical SOC processes. Currently, Splunk continues to help SOCs shift from a reactive to a proactive state with the following capabilities: Exposure Analytics, Detection Studio, Federated Search, and Agent-Driven SOC Expansion.