Live Webinar

Threat Modeling for the Autonomous Era

Key Takeaways from OpenAI's Hugging Face Incident

About This Webinar

A month after an OpenAI agent escaped a test sandbox and breached a third party's infrastructure, the incident is still drawing market and regulatory attention. The event exposed a problem many organizations have not fully addressed: agentic systems can act beyond their intended task, cross trust boundaries, and create risk faster than existing controls can respond.

AI regulation is coming. NIST is preparing new guidance, lawmakers are advancing agent-focused proposals, and industry frameworks are evolving quickly. But specific requirements are not yet in place, meaning organizations must define and implement their own approach to securing agentic systems.

This session will break down what went wrong in OpenAI's Hugging Face incident and examine how teams can use agentic threat modeling to identify AI risks before, during, or after deployment. Attendees will walk away with a practical guide for reviewing agentic architectures, defining controls, and strengthening governance.

Your Panel

• Ben Oster — Chief Product Officer, ThreatModeler (Moderator)
• John Steven — Founding Principal, Aedify Security
• Nikunj Nagalia — VP, Solutions Architecture, ThreatModeler