Cisco Systems has announced security innovations focused on the agentic AI ecosystem, where AI systems perform actions beyond answering queries. Unveiled at RSA Conference 2026, the solutions are intended to address AI security issues and support the adoption of agentic technologies.
The company is incorporating security measures into AI deployment through a combination of approaches, including establishing identities, applying Zero Trust Access controls, safeguarding agents, and expanding the capabilities of security operations centre (SOC) teams.
According to a Cisco survey, 85% of major enterprise customers are experimenting with AI agents, while 5% have deployed them in production. In response, Cisco is focusing on three areas: limiting unintended agent actions, protecting agents from external interference, and improving the detection and response to AI-related incidents.
AI agents require defined identities, assigned functions, and accountability to human managers. However, many organisations currently lack visibility into which agents are in use or how responsibility is assigned.
Cisco is extending its Zero Trust Access model to AI agents, introducing access controls supported by Duo IAM capabilities. Organisations can register agents and link them to human owners, enabling identity verification, activity tracking, and oversight of agent usage.
As organisations expand AI deployments, Cisco is also updating its AI Defense offering. This includes Cisco AI Defense: Explorer Edition, a self-service tool designed to test and assess AI agents before deployment within agent-based environments.
The offering includes features such as dynamic agent red teaming, security testing, and reporting, intended to help organisations evaluate risks and identify vulnerabilities.
Cisco is also incorporating AI capabilities into security operations through its Splunk platform. Tools such as exposure analytics and detection capabilities are designed to support continuous risk assessment and help SOC teams prioritise and respond to threats.
The use of specialised AI agents within security workflows is intended to automate certain processes and support faster response times, reflecting changes in how organisations manage evolving security challenges.
