The AI Agent Security Summit Returns to San Francisco: Meet the First Speakers

Since our last AI Agent Security Summit in 2025, the conversation around agentic AI security has shifted considerably, and a new set of challenges has moved from theoretical to pressing.

OpenClaw crossed 250,000 GitHub stars and introduced a new class of enterprise exposure. Agentic browsers entered the threat landscape. The OWASP Top 10 for Agentic Applications gave the industry its first globally peer-reviewed framework for assessing agent risk, and security teams quickly recognized how far their existing programs were from meeting it. The agentic attack surface has expanded, the stakes are higher, and the gap between adoption speed and security maturity is still very much open.

On May 27, the summit returns to the Commonwealth Club to address all of it, with a full-day agenda built around what's actually changed, what's still noise, and what security leaders need to act on now.

What to Expect at the AI Agent Security Summit

The AI Agent Security Summit is a community-led event for practitioners in the AI Security ecosystem, including CISOs, security architects, and security researchers who work at the intersection of AI agents and enterprise risk. Sessions are grounded in original research, live simulations, and real case studies from organizations already running agentic systems at scale. We are bringing back the beloved format of keynotes, sessions, lightning talks, panels, and more that will make the day engaging and interactive.

This year's agenda reflects where the threat landscape actually is today. You'll hear original research showing that prompt hardening achieved 100% prevention against moderate attacks and exactly 0% against sophisticated ones, with a deterministic infrastructure-level control as the only defense that held. You'll see how skill registries are being exploited through social proof manipulation, attackers defeating scan badges while delivering malware to thousands of users who assumed popular means safe. Lyft will share how they automated triage across more than 5,000 vulnerability findings using a production agent platform built around real infrastructure context. And a live supply chain lab will walk through real attacks, then show the same environment hardened with SBOM validation and provenance enforcement in real time.

The day also addresses the foundational questions security leaders are grappling with right now, including a panel unpacking AIUC-1, the first standard built specifically for agentic AI, a deep dive into why API gateway protection alone leaves enterprises exposed, and a comparative study of autonomous red teaming tools that shows clearly where today's testing approaches go blind. These are the conversations that don't happen at general security conferences, because the people in the room are the ones actually building, breaking, and defending these systems.

Meet the First Speakers of the AI Agent Security Summit

In keeping with Summit Tradition, the kickoff and wrap-up of the day will be presented by none other than Michael Bargury, Co-founder and CTO of Zenity, and joining him will be seventeen speakers across eleven sessions, with more to be announced in the coming weeks. Here’s a sneak peak at a few of the amazing sessions at this upcoming AI Agent Security Summit.

Travis McPeak, Cursor

A fork in the road for security: Security organizations have set new breach records year after year despite widespread acknowledgment that security matters. Travis examines whether AI is the tool that finally tips the balance toward defenders, or whether it accelerates the problem further, and what that means for how enterprises structure their security programs today.

Jim Reavis, CEO, Cloud Security Alliance (CSA)

Cybersecurity A.M. (After Mythos): The Mythos project has accelerated vulnerability discovery and exploit development at a pace that challenges long-standing assumptions about defense. Jim breaks down what has actually changed across the full cybersecurity program lifecycle, what hasn't, and where security leaders should concentrate their attention and investments now.

Jenn Gile, Co-Founder, OpenSourceMalware

When green doesn't mean go: How AI skill registries create false security: Drawing on original research into a sustained malicious skills campaign targeting ClawHub, Jenn breaks down how attackers exploited three compounding failure modes to deliver malware to thousands of users while appearing completely safe. Attendees will leave with a clear understanding of why static scan badges, download counts, and star ratings create a false sense of security, and what signals teams should actually use when evaluating skills for enterprise adoption.

AIUC-1: The First Agentic AI Standard (Panel)

Panelists:

• Chris Hughes, VP of Security Strategy, Zenity
• Rinki Sethi, CISO, Upwind Security
• Jim Reavis, CEO, Cloud Security Alliance
• Rajiv Dattani, Co-Founder, AIUC

Agentic systems introduce risks that static compliance models weren't built to handle, including autonomous decision-making, dynamic tool use, non-deterministic behavior, and machine-speed identity proliferation. This panel of practitioners and standards architects explores what AIUC-1 prescribes that existing frameworks miss, and how security leaders, builders, and buyers can use it to accelerate safe enterprise adoption.

Join Us on May 27, 2026, in San Francisco

The AI Agent Security Summit is free to attend. Sessions run throughout the day at the Commonwealth Club in San Francisco, with dedicated time for conversation and connection with speakers and fellow practitioners. If you're working at the intersection of AI agent security and enterprise risk, this event is built for you.

More speakers will be announced in the coming weeks.

Secure your spot now: Register for the AI Agent Security Summit