Rebecca PY Cisco introduced a suite of security solutions at RSA Conference 2026 to address risks in the rapidly growing agentic AI ecosystem. The innovations aim to enable enterprises to safely deploy AI agents at scale with Zero Trust controls and real-time threat response.
MALAYSIA, 26 MARCH 2026 – Cisco has announced a comprehensive set of security innovations designed to support the emerging agentic AI ecosystem, where software systems move beyond assisting tasks to actively executing them. The announcement was made at RSA Conference 2026, highlighting Cisco’s push to remove key barriers to enterprise adoption of AI agents.
As organizations increasingly explore agentic AI, Cisco revealed that while 85% of enterprises are experimenting with AI agents, only a small fraction have deployed them in production. Security concerns remain a major obstacle, particularly around identity, access control, and threat detection.
To address these challenges, Cisco is focusing on three core pillars: protecting the world from AI agents, protecting AI agents from external threats, and enabling rapid detection and response to AI-driven incidents.
A key highlight is the extension of Zero Trust Access to AI agents, allowing organizations to establish verified identities for agents and link them to accountable human owners. Through enhancements in Duo Identity and Access Management (IAM) and Cisco Secure Access, enterprises can gain visibility into agent activity, enforce strict access controls, and ensure agents operate within defined parameters.
Cisco is also strengthening defenses around AI systems with the launch of AI Defense: Explorer Edition, a self-service platform that enables organizations to test and secure AI models before deployment. The solution includes capabilities such as adversarial red teaming, prompt injection testing, and actionable security reporting, helping developers and security teams identify vulnerabilities early.
Further supporting secure development, Cisco introduced an Agent Runtime Software Development Kit (SDK), which embeds policy enforcement directly into AI workflows. The company also launched the LLM Security Leaderboard, offering transparency into model risks and resilience against malicious inputs.
In addition, Cisco unveiled DefenseClaw, an open framework designed to integrate security directly into AI development pipelines. By combining tools for scanning, verification, and asset management, the framework enables organizations to deploy AI agents securely while maintaining operational efficiency.
On the operations front, Cisco is enhancing security operations centers (SOC) through its Splunk platform by integrating AI-driven capabilities that shift security from reactive to proactive. New features such as Exposure Analytics, Detection Studio, and Federated Search provide real-time visibility and streamlined threat detection. The introduction of specialized AI agents within the SOC further automates tasks such as threat analysis, incident response, and workflow execution.
These innovations reflect Cisco’s broader vision of embedding security into the foundation of the AI-driven economy. By enabling organizations to manage risks while scaling AI adoption, the company aims to unlock the full potential of agentic AI as a transformative force in enterprise operations.