What I'm Watching at Gartner Security Summit 2026
by Beth Barach, VP of PM//7 min read/
The Gartner Security & Risk Management Summit opens June 1 in National Harbor, and for the first time in years, the conference agenda feels genuinely urgent, in the way that happens when the threat environment shifts faster than the industry's ability to describe it.
Between Anthropic's release of Claude Mythos Preview in April and OpenAI's Daybreak announcement just days ago, the security community is walking into this event carrying a shared, unspoken question: does everything we built still hold?
I have gone through the agenda and although it was challenging to choose, I have planned to attend the sessions I think will help evolve my point of view about exposure management. Three themes keep surfacing that I think will define the real conversations that will happen during the summit.
AI Is Not an Add-On Anymore. It Is the Architecture.
Several sessions this year are built around a premise that would have felt premature two years ago: AI isn't a tool you layer onto your security program. For a growing number of organizations, it is becoming the architecture itself. The question that follows is harder than it sounds. When AI drives discovery, prioritization, and response, what do the operational systems underneath it actually need to look like?
This matters right now because Mythos answered the offensive half of that question in a way no one expected. Anthropic's model autonomously found thousands of critical zero-day vulnerabilities across every major operating system and browser, without human guidance, in weeks. It didn't scan. It reasoned. It read code, formed hypotheses, tested them against running binaries, and built working exploits. OpenAI's Daybreak followed with a different angle: use AI to embed security into the development lifecycle from the start, not just find and fix what already exists.
That’s two major AI labs, within weeks of each other, placing aggressive bets on AI-native security. The offense is now operating at a speed and scale the industry has never had to defend against. The sessions I'm most interested in on this theme aren't asking whether AI belongs in security. That debate is over. They're asking what AI-native security actually requires of the program underneath it. Discovery without context produces noise. Autonomy without a verified data foundation produces risk. The architecture conversation is really a data conversation.
Exposure Management Is Moving from Framework to Operational Requirement
Continuous threat exposure management (CTEM) has been a Gartner framework for several years now. What I would like to hear at this summit is a true assessment of where organizations actually are in the journey from understanding the framework to running it as a live operational discipline.
Mythos accelerated that timeline whether the market was ready or not. A 2025 report found that over 45% of discovered vulnerabilities in large organizations remain unpatched after 12 months. That number existed before AI could weaponize a disclosed CVE in the window between publication and patch deployment.
What that means operationally is that exposure management is no longer a program maturity aspiration. It is a response capability. The organizations best positioned to absorb AI-speed discovery are the ones that already have continuous visibility across their full exposure surface. They apply business context to every finding. They route remediation to the right owner with SLA attached. That operational foundation is what keeps programs from collapsing under the volume; that's vulnerability prioritization working the way it's supposed to.
Agentic AI Needs Governance Before It Needs More Capability
This is the theme I'm most curious to see addressed directly. The push toward agentic AI in security, AI agents that take autonomous action on findings, route remediation, execute fixes, is real and accelerating. However, deploying an AI agent capable of autonomous security analysis means introducing a new category of actor into your environment. Who authorized it? What data and systems can it reach? What actions can it take without human review? How do you audit what it did after the fact?
More AI capability does not mean more security. What makes autonomous action trustworthy is the data foundation it operates on. Clean, deduplicated, attributed findings with verified ownership and business context attached. Compensating controls mapped against every attack path. Prioritization grounded in actual environment context, not just CVSS scores. Without that foundation, agentic AI doesn't accelerate your program. It accelerates your noise.
And as AI takes on more offensive and defensive capability, the skills that matter most for defenders are shifting. Finding vulnerabilities matters less. Contextualizing them, prioritizing them, and governing the systems that act on them matters more. That not only changes the skills needed but also the tools being used.
What I'm Expecting, and What I'm Interested In Learning More About
This summit arrives at a moment when the industry has the opportunity to move beyond debating whether AI belongs in security and into the harder, more valuable work of figuring out what AI-native security actually requires. It means treating agentic AI governance not as a future consideration but as a prerequisite for autonomous action that is safe to run today. I am interested in learning more about how programs are absorbing AI-speed discovery, where prioritization is breaking down, and what it actually looks like to route remediation to verified owners at scale.
Come Meet the Brinqa Team
If you're attending the summit, come find us at the Brinqa booth in the Exhibit Showcase.
We're also presenting Monday, June 8 at 6:30 PM EDT: "A CISO's Guide to Real Risk Reduction in Exposure Management" with Jay Klauser at Theater 4.
The Gartner Security & Risk Management Summit runs June 1-3, 2026 in National Harbor, MD.
