Cisco on March 23, local time, unveiled a large number of AI agent security solutions ahead of the RSAC 2026 security conference, which will be held in San Francisco, the United States.

According to a report by Techzine, Cisco introduced zero-trust access controls for AI agents, a new version of its AI model red-teaming tool, and AI agents tailored for security operations centers (SOC).

Tom Gillis (톰 길리스), senior vice president and general manager of Cisco's Infrastructure and Security Group, said, "For decades we have managed access for people under the principle of least privilege. AI agents have human-level privileges but printer-level judgement."

That means they have extensive privileges but poor judgement. This is the key reason AI agents become a security threat.

Cisco appears to be trying to address this issue using the zero-trust security concept.

Cisco registers AI agents in its identity and access management (IAM) solution, Cisco Duo, and links them to the responsible employee. It assigns each agent a unique identity so it can track who operates which agent. It grants agents only granular access rights limited to the tasks in charge and the time needed.

Such zero-trust access control will be built into Cisco's security service edge (SSE) solution, Cisco Secure Access. It places a Model Context Protocol (MCP) proxy within the SSE and routes all agent-to-agent communications and agent-to-external tool communications through the proxy. Gillis explained, "We treat MCP traffic the same as HTTP traffic."

AI agent security cannot be achieved by protecting the agent itself alone. If the AI model used to run the agent is vulnerable or compromised, the agent is also put at risk. With this in mind, Cisco released a new version of AI Defense called "AI Defense: Explorer Edition."

"AI Defense: Explorer Edition" supports developers, application security teams and security researchers so they can conduct red-team tests on AI models and applications themselves before deployment to production environments.

Cisco is also deploying AI agents in SOCs. The new agents Cisco is releasing are a detection builder, standard operating procedure (SOP), triage, malware threat reverse analysis, guided response and automation builder. Of these, the malware threat reverse analysis agent has been formally released, and the rest will be rolled out in stages by June this year. Splunk Federated Search will be used for large-scale data processing performed by AI agents and models.