Vulnerability management has been losing ground for years, and AI is about to widen the gap on both sides of the equation. Attackers are using AI tools to discover and weaponize flaws faster than disclosure programs can keep up, while defenders are still triaging findings through a patchwork of scanners, ticketing systems and manual review that wasn’t built for this pace. Treating security as a separate workstream from the rest of the AI-era infrastructure stack is what keeps that gap open.

Peter Bailey of Cisco sits down with Alan Shimel at Cisco Live to make the case that security has to be designed into the same fabric as networking, observability and AI services — not bolted on after the fact. Bailey’s argument is that the convergence of Splunk data, Cisco’s security controls and Cloud Control under a single AI-driven layer is what finally makes it possible to see vulnerabilities, exploitability and business context inside one view instead of three disconnected ones.

Bailey gets into the practical mechanics. AI-assisted discovery surfaces issues across sprawling estates faster than human teams can sort through them, but the value only shows up if remediation moves at the same speed — automated patching at scale, compensating controls that shield legacy systems while fixes are staged, and prioritization tied to actual exploit likelihood rather than CVSS theater. Reducing tool sprawl matters here for a practical reason: every additional console adds latency between discovery and action.

The bigger thread is industry-wide urgency. Software quality and the long tail of unpatched systems were already a problem; agentic attackers will not wait for slow disclosure cycles or quarterly maintenance windows to catch up. Bailey’s view is that the security teams getting ahead are the ones treating AI as core infrastructure for defense — using it to compress the time from vulnerability discovery to remediation, instead of waiting for it to be used against them first.