Most organizations believe they have visibility into all of their identities. They do not, and many don’t even come close. For every human account within your organization, there are an average 80 non-human identities quietly operating in the background.
These include service accounts, API tokens, CI/CD bots, AI Agents, and OAuth credentials that were provisioned, forgotten, and never reviewed. Add orphaned human accounts, over-privileged contractors, and shadow SaaS platforms operating entirely outside your identity governance program, and the risk grows exponentially. Your identity blind spots aren’t just edge cases; they are a risk throughout your organization.
This presentation examines the five categories of identity security blind spots that threat actors exploit before defenders ever see the alert: non-human identity sprawl, human identity failures, cloud and shadow access, detection and governance gaps, and the absence of integrated identity providers and privileged access management (PAM) solutions.
It is not a matter of if a threat actor will find your blind spots, but rather whether you find and remediate them first.
*Sponsored by Beyond Trust
Virtual
SOC teams face unprecedented threats in today's fast-evolving landscape. In this session, learn how Google Security Operations and Wiz Defend work together to power detection and response for the agentic era. Security practitioners will discover practical use cases and actionable strategies to scale operations using agents, gain unified visibility across attack surfaces, and see these capabilities in action during a live demo.
*Sponsored by Google
Virtual
A financial advisor's AI assistant, asked to summarize market news, follows an instruction hidden in an article and emails another client's portfolio outside the firm, and every step uses access the advisor legitimately holds. Employee-level controls answer whether a person may act, but agents also need an answer to a second question: should this agent do this, on this task?
This session shows how to add that missing layer as lanes: an agent's job expressed as policy, written in plain English with zero-shot intent and entity detection, and enforced at each step of the agent's loop. A live demo shows the same action getting opposite outcomes in two lanes.
*Sponsored by Witness AI
Virtual
Virtual
Mythos-class AI models have made attackers autonomous, collapsing the window from disclosure to exploitation from weeks to hours and leaving every security leader with the same question: are we Mythos-ready? Getting there takes two shifts: validation must become adaptive, triggered by threat and environment changes instead of the calendar, and decisions must rest on proof of what is actually exploitable.
This session shows an agentic approach where specialist AI agents, working under human command with tunable autonomy, sense change, prove exploitability, test security controls, and mobilize remediation end to end. Attendees will leave with a practical blueprint for human-on-the-loop Mythos readiness: every change validated within hours, and the evidence feeding straight into your CTEM program.
*Sponsored by Picus
Virtual
Across the modern enterprise, AI is being adopted faster than it can be secured. AI systems and agents, from sanctioned copilots to unsanctioned tools and custom-built agents are introducing risk that traditional security tools cannot see.
This session examines how behavioral security closes that gap, giving security teams visibility into how AI is actually behaving across the enterprise. The result: security becomes an enabler of AI innovation rather than a barrier, allowing organizations to adopt AI with confidence.
*Sponsored by Darktrace
Virtual
As AI agents become increasingly autonomous, traditional approaches to identity and data security are struggling to keep pace. Agents inherit permissions, access sensitive information, and interact across systems in ways that create new security blind spots.
This session explores why securing AI requires a unified understanding of agents, identities, permissions, and data. We will demonstrate how Zscaler's graph-based approach maps these relationships to provide visibility into data exposure, transitive access, and AI-driven risks.
Through real-world examples and demonstrations, we will explore how organizations can enforce least privilege, trace data flows, and contain threats. Attendees will learn how to reduce risk while accelerating secure AI adoption.
*Sponsored by Zscaler
Virtual
Virtual
AI agents have deleted production databases, wiped developer files, and tried to hide their actions from human reviewers. We put guardrails around them, so why are they doing these things?
These examples sound extreme until you understand the logic behind them.
Join Cato Networks' Etay Maor, VP Threat Intelligence and Founder, Cato CTRL, and Ryan Ashcraft, AI Security Engineer, as they examine why AI agents do the things they do, how attackers manipulate their behavior through techniques like prompt injection, and what those incidents tell us about what's needed from AI Security. The AI tools employees use, the AI applications developers build, and the AI agents operating across enterprises all present different risks, but they all need the same foundation to stay controlled and governed: visibility, posture, and runtime protection.
Through a live demonstration of Cato AI Security, you'll see how one solution with one context and one control helps security teams see, understand, and secure every AI interaction.
*Sponsored by Cato
Virtual
AI coding agents now generate code faster than security teams can review it, and they install dependencies, invoke MCP tools, and execute commands with developer-level permissions. Faros AI's 2026 engineering report found that median pull request review time rose 441.5% with high AI adoption, while 31.3% more pull requests were merged without any review.
We’ll demo how AURI by Endor Labs adds a security harness to AI coding tools developers already use like Claude Code, Cursor, and Codex to: Govern agent actions by inventorying every agent, model, and skill, blocking destructive commands, and logging every action to an audit trail. Generate secure code by default with real-time context from AI SAST and secrets detection. Detect and fix what slips through with AI SAST and automated remediation agents.
*Sponsored by Endor Labs
Virtual
AI coding agents are moving from pilot to production, making decisions and taking actions with real financial and legal consequences, often with little oversight.
This session unpacks that risk: non-technical teams now directing powerful agents, daily malware attacks outpacing even the most sophisticated security teams, and code volumes growing faster than human review can scale. Join us to learn how to contain agent risk without AI-scale spend.
*Sponsored by Semgrep
Virtual
Virtual
AI has changed the rules. Security must now account for what employees share with AI, how AI interprets information, and what agents do next, across applications, interfaces, and workflows.
In this session, we'll show what it takes to rewrite the AI security playbook: understanding content, context, and intent at the interaction level, translating business requirements into enforceable policies, and applying precise, real-time interventions that protect sensitive information without disrupting legitimate AI use. All through your existing infrastructure.
*Sponsored by Lumia Security
Virtual
