Umesh Mahajan & Prashant Gandhi, Broadcom | RSAC 2026
summary
Broadcom argues agentic AI lets adversaries crawl millions of lines of open‑source code at machine speed, making perimeter defenses dangerously insufficient, so enforcement sits at the hypervisor alongside workloads rather than bolted on. The stack claims 6X SSL performance gains in Avi, adds post‑quantum crypto to prevent “harvest now, decrypt later,” and uses Model Context Protocol intelligence to spot agentic AI traffic. A prescriptive phased zero trust model targets the ~75% who stall, customers cut support tickets by up to 90%, and Broadcom IT stress‑tests the stack to protect semiconductor IP.
AI-written summary · about
chapters (19)
- 0:00Introduction to RSAC 2026
- 0:49Broadcom’s security vision
- 2:11CISO priorities for AI security
- 3:39AI-defined security platforms
- 5:07Seamless infrastructure integration
- 7:21Broadcom’s engineering roots
- 8:04Escalating AI-driven threats
- 10:30Agentic AI security challenges
- 11:33Open source malware risks
- 12:18Zero trust lateral security
- 13:41Operationalizing zero trust
- 15:09Prescriptive threat prevention
- 16:55Embedding security deeply
- 18:39AI in security detection
- 19:01Eating our own dog food
- 20:06Avi load balancer innovation
- 23:04Agentic AI attack surface concerns
- 24:53Shadow AI governance risks
- 26:20Securing AI workloads
