Thomas Prigg, AI Incident Response Team lead, explores a new approach to AI threat detection using behavioral drift analysis. By applying data science techniques to AI agent activity, defenders can identify prompt injection, poisoning, malicious adaptation, and other threats earlier than traditional security methods.
Join Ads Dawson of Dreadnode and Mike Takahashi of Zenity — ranked among the top bug bounty hunters on HackerOne and Mozilla's AI programs — as they break down a year's worth of data exfiltration bugs found across production AI applications, showing how the same primitive behind XSS now lets attackers steal PII, business data, and private messages through AI agents.
Summit: August 14 @ 10:00 am – 4:30 pm EDT How do you secure a future that is already here? AI is changing how organizations operate, how defenders protect systems, and how attackers identify and exploit opportunities. New capabilities, new risks, and new questions are ...
Join Ethan Robish, Black Hills Infosec former pentester and current SOC architect, for a technical session on the scaffolding, guardrails, and processes that turn agentic coding from a novelty into a production-grade workflow.
In this session, Hayden Covington, Associate Director of Security Operations at the BHIS SOC, will walk through how to design custom agents for real security operations work: triage support, detection engineering, enrichment workflows, reporting, research, and review-heavy operational tasks. The focus is not magic prompts or over-engineered instructions; it is the engineering work behind useful agents. Powerful agents have clear roles, bounded authority, durable memory, tool access, handoff points, review gates, and failure modes that operators can actually relate to.
Join Molly Correia, cybersecurity consultant, as she turns a virtual talk into a live experiment: you'll prompt your preferred AI model — ChatGPT, Claude, Gemini, Copilot, or whichever model your team relies on — to analyze a real-world CVE, then compare results against authoritative sources to see exactly where the models hallucinate facts, recommendations, and unsupported conclusions.
This panel moderated by Kip Boyle brings together two hands-on AI security pros, a privacy and AI attorney, and a cyber-insurance expert to work through where the guardrails actually belong and who owns them once they're set.
Do you dread writing incident response documentation? Gerard Johansen of IR Proactive shows how LLMs can quickly create, refine, and validate incident response policies, plans, workflows, and communications playbooks.
Join Faan Rossouw (aionsec.ai) for a 25-minute, demo-driven session on a blind spot in every security system that puts an LLM agent in the analysis loop: the data the agent reads is partly authored by the adversary, which turns your own telemetry into a potential prompt-injection channel.
Join Alicia Cuestas, CEO of Open-Sec and offensive security professional, as she demonstrates how to exploit agentic AI systems by treating the LLM itself as a vulnerable jump-box. Using indirect prompt injection and tool-use manipulation, she'll show you how attackers leak sensitive data, move laterally across internal APIs, and bypass system-level guardrails.
Join Matt Franz, AI Security Lead at Bespin Global, as he shows you how to vibe-code a custom MCP Server for your SIEM in Go, whether or not your vendor already ships one.
