This interactive tabletop exercise challenges participants to respond to a sophisticated, multi-vector incident that targets their AI infrastructure through unexpected attack surfaces. Your customers look to you to lead in times of disruption and crisis. This exercise will challenge traditional thinking and disrupt the approach to incident and crisis management.

Your DevOps lead just authenticated via FIDO2 from a compliant corporate laptop. The IdP says "Authentication Success," but their underlying AWS API calls are silently staging a database snapshot. This is the post-authentication blind spot: the vulnerability that occurs when organizations treat authentication as a point-in-time gate rather than a continuous behavior loop. This is a tabletop simulation of an active session-hijack riding a valid SSO token. Participants will look past the initial login gate to explore the gaps between your Identity Provider and cloud infrastructure. Collaborate with your peers to debate who owns post-auth behavior (IAM vs. SOC), analyze the gaps in standard SIEM logic, and outline practical rules for detecting lateral movement inside authorized accounts.

Many organizations have strong visibility into identity and access, but far less confidence in their ability to actually control it. This workshop will focus on how security teams are managing access in practice across SaaS, non-human identities, and constantly changing environments. We will compare what is working, where traditional approaches fall short, and how teams are evolving beyond static roles and periodic reviews to achieve more reliable control over access.

Your workforce is already using Claude, ChatGPT and other AI assistants, and development teams are shifting coding, testing and QA work to autonomous agents that connect to APIs, MCP servers, SaaS applications, cloud infrastructure and production data. Authentication is only the beginning. In this hands-on workshop, we'll trace an agent request from its originator to the sensitive resource it's trying to reach, mapping the controls along that chain (network controls, API security, MCP gateways, agent identity, model guardrails and runtime authorization) and examining what each one can actually see and stop. The goal: answer what your identity and security stack needs to look like when agents act on behalf of people, NHIs and other agents. You'll leave with a framework for evaluating your agent security stack, identifying gaps and deciding where authentication, connectivity controls and runtime authorization belong as agentic adoption expands.

The cybersecurity industry has never had more engineering talent, more AI capability, or more investment. Yet many of our biggest operational problems remain largely unchanged. We keep innovating around detection, summarization, copilots, and automation, while organizations still struggle with ownership, accountability, authority, remediation, and decision-making, the problems that actually determine whether security outcomes improve. Is this the natural evolution of technology, or are we optimizing for problems that are easier to demonstrate rather than the ones that are harder to solve? In this interactive discussion, we'll explore questions such as why some cybersecurity problems attract enormous innovation while others go untouched, whether we are optimizing for technical elegance or operational impact, what we would build differently if enterprise adoption, not product demos, were the measure of success, where the next decade of AI investment should actually go, and which problems are waiting for someone willing to solve the uncomfortable parts of cybersecurity.

Practitioner To Be Announced

AI has made it easier than ever for security and compliance teams to build their own automation, and the tools have improved dramatically in just the past year. In this roundtable, Murat opens with a few honest stories from his own builds: what worked, what broke, and what he’d do differently with today’s tools. Then the room takes over. Participants share their own experiences, stress-test them against real-world pressures, and assemble a shared recipe for when building makes sense, what keeps homegrown automation alive, and where human judgment has to stay in the loop.

Most security teams believe AI could do their work better, but getting from a promising prototype to something accurate, consistent, and affordable at scale is where most efforts stall. Participants dig into the hardest problems along that path: building a context layer accurate enough to feed the agents, choosing between one large prompt and many specialized ones, selecting the right model for each task, and designing an architecture that balances accuracy, efficiency, and cost. The session compares what worked, what broke, and what security leaders should ask before trusting an agentic system in production.

NPM Imposters is a card game where players assume the role of an organization's Head of Application Security to protect their software supply chain from hidden malware. To win, players must root out malicious code hidden within both immediate and transitive dependencies, aiming to end up with five face-up package cards that are completely free of malware.

As attack surfaces grow and automated reconnaissance gets faster, discovery of exposed assets is increasingly a given. When being found is inevitable, does "likelihood" still help leaders separate the risks that matter most? This workshop is a forum for leaders to compare notes on prioritizing exposure, communicating risk, and allocating resources under that assumption. The goal is not another universal scoring model, but practical ways to make risk measurement useful. If every meaningful exposure will be found, risk is less about whether contact occurs and more about what an attacker can do next, how far they can go, and how well the organization can respond. Start with a real-world scenario: assume discovery. A business-critical service has an externally discoverable weakness. Should discovery still be scored probabilistically, or treated as a given? Add exploitability and consequence. Reveal compensating controls, identity paths, data sensitivity, and blast radius. Groups decide what actually changes the priority. Add organizational response capacity. Reveal differences in detection, containment, recovery, and control confidence. Do two companies with the same exposure carry the same risk? Synthesize. Each group answers: "If likelihood is no longer the primary discriminator, which three measures should replace or supplement it?"