A recent Cisco survey highlights a significant gap: while 85% of enterprises are piloting AI agents, just 5% have deployed these agents into full production, largely due to trust concerns. Cisco’s President Jeetu Patel identifies trust as the crucial factor distinguishing market leaders from those at risk of failure. AI agents resemble intelligent yet inexperienced teenagers needing strict oversight to prevent costly mistakes. At RSA Conference 2026, Cisco announced new security tools like AI Defense Explorer Edition and Defense Claw, addressing three key areas: protecting agents, safeguarding the environment from agents, and detecting threats rapidly. Nvidia’s OpenShell integration accelerates secure deployment by automating security policies at launch. Patel also revealed a bold plan to have 70% of Cisco’s products built entirely by AI by the end of 2027, signaling a cultural shift in engineering. Five strategic “moats” such as sustained speed and trusted delegation guide enterprises toward safer, more efficient AI adoption. Experts warn that distinguishing agent-initiated actions from human ones remains a challenge, urging enterprises to enhance telemetry alongside identity controls to close security gaps. To move forward, security teams should audit trust barriers, red-team AI workflows, map delegation, establish behavior baselines, and improve logging capabilities to ensure safe AI agent operation.
Back
