Cisco is combining its network telemetry with OpenAI models to help enterprise security teams detect and respond to threats at machine speed. The effort comes as attackers use AI to automate attacks and develop exploits more quickly, according to the companies’ account of the work.
Cisco’s approach distributes security controls across microservices, containers and clusters rather than relying on a single perimeter. Cisco’s Patel says controls can be placed in front of individual services and Kubernetes environments, with security integrated into the network. Application-level rules can limit services to necessary connections, while selective isolation can contain affected systems without interrupting services that are not involved.
The companies say OpenAI models can help Cisco correlate network signals at scale. Patel notes that separate signals may appear harmless but reveal a significant pattern when considered together. This network-wide analysis is intended to give security teams more information to guide their response.
Patel identifies the time between disclosure of a vulnerability and an organization patching it as a key challenge. He says AI can help defenders investigate more threats and reduce the time required for investigation, response and remediation. Cisco also says initiatives including Daybreak and GPT-6 Astra have improved its own security posture and that of its customers, and can be incorporated into existing vulnerability-scanning workflows.
The work also addresses risks from increasingly autonomous agents. Patel says teams need to combine security telemetry with observability data to distinguish malicious activity from an agent behaving unexpectedly while pursuing a legitimate goal. He outlines three priorities: protecting agents from attacks, detecting and interrupting unintended behavior, and operating in near real time.
Through the Daybreak Defense Network, Cisco and OpenAI are exploring ways to help security teams triage exposure faster and apply controls in Cisco Cloud Control. OpenAI’s stated enterprise data protections include zero data retention, strict customer-data boundaries and verifiable isolation.
Sources: OpenAI News.