Most compliance programs still run on a calendar: evidence gathered in a sprint, controls verified once, and a snapshot defended until the next audit. In this session, RegScale CISO Dale Hoak shares his view of where legacy GRC breaks down, while Senior Solutions Engineer James Sumka shows what continuous controls monitoring looks like in practice, and Product Solutions Owner Charlie Jones previews what the team is building to solve those gaps and where RegScale is headed.
*Sponsored by RegScale
Virtual
AI attackers do not behave like human intruders. They move quickly, operate continuously, and may interact with decoys even after identifying them as suspicious. In one study of 21 language models, AI attackers took the bait more often than human attackers, and many continued despite recognizing a possible trap.
This session examines how security teams can use deception to detect machine-speed attacks and validate whether their existing controls respond to real attacker behavior. We will explain how NodeZero uses autonomous attack-path validation to identify where an attacker can actually move, then deploys Tripwires and AD Tripwires along those proven paths.
*Sponsored by Horizon3
Virtual
Modern attacks start at the edge, and the widest gap in most environments is the space between the tools people already run. AI widened that gap as agents act continuously, at machine speed, on devices nobody registered. Every one of them has to resolve a name before it does anything else, which makes DNS the earliest place to see the behavior and the earliest place to stop it.
*Sponsored by DNSFilter
Virtual
Virtual
Abliterating an open-weight model normally means editing and redistributing tens or hundreds of gigabytes of weights.
Weightless takes a different approach: extract the behavioral direction and apply it at inference time, leaving the base checkpoint byte-identical. I’ll show how projective activation steering can suppress refusal using compact GLP artifacts (for example, a 478 KB vector for DeepSeek V4 Flash) and how the same intervention can be expressed as a rank-1 LoRA on compatible dense architectures. The underlying runtime method is not limited to dense models: it also applies to mixture-of-experts architectures such as GLM, where activation-space intervention is the more natural deployment mechanism.
We’ll cover real measurements across DeepSeek, Qwen, and GLM models on DGX Spark, including refusal reduction, capability and benign-behavior gates, quantization behavior, deployment constraints, and the cases where steering breaks.
*Sponsored by Tolmo
Virtual
AI is spreading across the enterprise (from public tools and embedded SaaS capabilities to copilots and autonomous agents), creating a new security challenge: how do you govern an AI environment you can’t fully see? In this session, Karan will explore how familiar security principles like visibility, least privilege, monitoring and policy enforcement need to evolve as AI moves from answering questions to taking action. He’ll show how organizations can discover sanctioned and unsanctioned AI, understand what systems and data it can access, and establish controls over what it is allowed to do. Attendees will leave with a practical framework for bridging traditional security and AI governance, including a demonstration of Perimind Observe.
*Sponsored by Cinchy
Virtual
Autonomous AI coding agents have the same access rights as senior engineers, but rapid adoption often outpaces traditional security governance. Join this session to learn how Palo Alto Networks Prisma® AIRS™ delivers end-to-end visibility and control across the entire AI software development lifecycle. Discover how to protect every interaction—from initial prompt to final execution—while keeping developer productivity high.
*Sponsored by Palo Alto Networks
Virtual
Virtual
Your company wants AI transformation to accelerate, and your workforce is losing patience. Rather than wait for approval, employees are running agents, connecting MCPs, and authorizing risky permissions, leaving you with shadow AI and new threat vectors. To enable AI securely, you need to make the safe path the golden path. This session shows you how Runlayer's AI control plane delivers AI enablement, security, and governance in one platform, so you can accelerate AI transformation with security and control built-in.
*Sponsored by Runlayer
Virtual
Your next customer might not be human. AI agents are starting to search, compare and buy on behalf of real people, while attackers use the same automation to scrape, scalp and abuse. Matt Wilmot, Global CISO of Merlin Entertainments (LEGOLAND®, Madame Tussauds, SEA LIFE), joins Huskeys co-founder and CEO Itai Gafni to share what he learned running security at the edge of a global ticketing business. Families, partners, AI shopping agents and scalper bots all arrive at the same front door, and at the edge they can look the same. If your edge can’t tell them apart, security either loses you sales or lets attackers in, and stops protecting revenue and starts costing it.
*Sponsored by Huskeys
Virtual
The rapid enterprise adoption of generative AI and agentic architectures has fundamentally rewritten the cyber threat equation. Attackers are using AI models and automated offensive tooling to scan, chain, and weaponize cloud and application vulnerabilities at machine speed—compressing the window between vulnerability disclosure and active exploitation from weeks to mere hours. Meanwhile, enterprise security teams remain constrained by fragmented scanning tools, manual triage, and disconnected developer workflows. In this focused 15-minute demo, discover how Wiz enables cloud security leaders to establish an AI Threat Readiness posture across the entire build-to-runtime lifecycle.
*Sponsored by Wiz
Virtual
