November Meetup

Join us for the November Bay Area OWASP Meetup, proudly sponsored by Endor Labs.
Spend the evening with fellow security professionals for talks, networking, and discussion on how AI agents are changing both offense and defense.
Thank you very much Endor Labs for hosting us and providing the space, food, and drinks.

Agenda
5:00 PM : Doors open, Networking, food and drinks
5.30 PM :- Chapter introduction
5.45-6.30 PM :- The Coverage Failure in Autonomous AI Pentesting
6.30-7.15 PM :- Agent Security with Decision Models
7.15-8.00 PM :- [Talk 3 title]

Talk 1: The Coverage Failure in Autonomous AI Pentesting
Description: Autonomous pentesting agents often sacrifice breadth for depth, exhausting their budgets on a few promising leads while leaving much of the application untested.
We present a scheduler-driven multi-agent architecture that balances exploration and exploitation using a continuously updated coverage map tracking every tested surface, action, and result.
We’ll dive into recursive delegation, multi-user browser isolation, independent finding verification, and scoped agent capabilities. Finally, we’ll share benchmarks against existing AI pentesting tools and discuss the challenges of evaluating recursive multi-agent systems.

Speaker: Yacine Souam
Yacine is an R&D Engineer at Escape (YCW23), working on building AI Agents for automated pentesting.
He’s also a speaker at top security conferences, podcasts guest and co-author at OWASP AI Exchange.

Talk 2: Agent Security with Decision Models
Description: This talk covers securing AI coding agents by replacing rules with judgment. Rules judge the command, not its origin: the same command is routine when you type it and an attack when a README plants it. I’ll walk through asc-guard, which hooks into a coding agent and asks a decision model if each risky tool call is harmful, user-requested, or from untrusted content, then allows, asks or denies. I’ll share test results against real attacks, where it fails, and why “whose idea was this?” beats “is this dangerous?” Humans stay in the loop, but only where it matters.

Speaker: Srajan Gupta
I’m a Security Enthusiast with a decade of experience building security into the foundations of products and platforms - identity, boundaries, safe defaults, and fast feedback loops, in high-scale and AI-native environments.

Talk 3: TBD