Cisco taps OpenAI's Daybreak, GPT-6 Astra for machine-speed defense

Cisco's security argument starts from a shift Patel describes away from a single perimeter toward controls spread across microservices, containers, and clusters. That architecture is what makes network telemetry valuable: signals that look harmless alone can form a meaningful pattern when correlated across the network, which is the role Patel assigns to OpenAI's models.

The backdrop Patel cites is the gap between how fast attackers exploit a vulnerability and how long organizations take to patch it. His claim is that AI narrows that gap, and he points to Daybreak and GPT-6 Astra as step-function improvements rather than incremental ones. Adoption is framed as incremental, however — Patel says the capabilities fit existing scanning workflows.

A second thread is the agentic world, where security teams must tell malicious interference apart from an agent pursuing a legitimate goal in an unintended way. Patel's three priorities there — protect agents, detect and intercept unintended behavior, and operate in near real time — are guiding work through the Daybreak Defense Network and Cisco Cloud Control.