Analysis
Microsoft security executive Aarti Borkar argues that chief information security officers now have two connected responsibilities: govern an enterprise using agents, and adapt defenses as attackers use AI. Her October 28 post places identity and data protection at the center of both. Builders, she writes, should involve security teams before connecting an agent to company systems.
The practical recommendations start with knowing which agents exist, who owns them, what they can reach and how they behave. Borkar recommends governed identities instead of personal or shared credentials where supported and appropriate, with access limited to the task. She also calls for protected, classified data and combined signals across identity, endpoints, applications and cloud environments.
Our reading: an agent inventory becomes useful when someone can act on it during an incident. A name in a catalog is only a beginning. An operator needs to connect an observed action to the responsible agent, identify the permissions involved and stop further activity. The relevant acceptance test is whether that chain still works when the original builder is unavailable.
There is also an organizational question behind the technical controls. If a team can create an agent faster than it can establish ownership, security review will accumulate after deployment. A sensible implementation would make identity, a named owner and a revocation path part of the creation process. Microsoft’s post offers vendor guidance, rather than measured evidence that any particular organization has achieved that coverage.
Source: The CISO’s evolving role: Protecting and defending in the agentic era · 2026-10-28
Why it matters
Agent oversight becomes operational when an observed action leads to a known owner and a usable revocation path.