Agentic AI network security is the Black Hat conversation of the year. Doug Merritt, CEO of Aviatrix, joins Alan Shimel to unpack the Mythos escape, why agents behave more like humans than superintelligence, and how enterprises can bubble wrap every workload at the network layer.

About Doug Merritt

Doug ran Splunk for a decade and now leads Aviatrix, the network layer specialist for hybrid and multi cloud. Consequently, he brings pattern recognition from two of the most defining eras in enterprise software.

Inside agentic AI network security

Customers used to treat network security as a data center problem. As a result, cloud native and agent driven traffic slipped past legacy controls and gave attackers a fresh runway. Meanwhile, the Mythos incident showed that even OpenAI can miss a JFrog Artifactory door hiding in a sandbox.

In addition, Doug argues the fix is not more panic tools but a return to the three fundamentals: identity, endpoint and network. Therefore, teams need bubble wrap around every database, Kubernetes pod and serverless function, plus real egress controls on every agent path.

Why this matters now

Meanwhile, Kimi K3 and other open weight models are matching Mythos and shipping to anyone. Consequently, shadow AI is real, agent identity will dominate Black Hat and citizen developers will keep spinning up unsupervised workloads inside the enterprise.

Explore more cybersecurity coverage and the latest TechStrong TV interviews. Furthermore, Doug previews the Aviatrix approach to network level containment across cloud, edge and data center, plus what to look for at Black Hat next week.

For more information please visit aviatrix.com