Roundtable: Securing AI Against Adversarial Attacks [RSVP REQUIRED]

This event requires an RSVP; to attend, submit form here

AI systems are being attacked in ways that traditional security controls weren't designed to catch. Prompt injection rewrites model behavior at runtime. Data poisoning corrupts outputs at the training or fine-tuning layer. Model extraction lets adversaries reconstruct proprietary systems from query responses alone. These aren't theoretical---they're documented attack patterns showing up in production environments right now, against systems that passed every pre-deployment security review.


This roundtable puts practitioners in the room to work through what hardening actually looks like against each of these vectors: what defensive controls exist, where they break down, and what the architectural decisions are that make AI systems either resilient or brittle under adversarial conditions. Particular focus on middleware security layers---the governance and control surface between an organization's data and the models consuming it---which is where most enterprises have the widest gap between what they've deployed and what they've secured.


Rob T. Lee will facilitate. The format is peer exchange: what organizations in the room are actually deploying, where they're stuck, and what's working. Participants leave with a benchmarked view of the current defensive landscape and specific hardening actions they can take back to their teams.