[Virtual] "Prove It Wasn't the AI": A Live Agent Incident Post-Mortem
Register with organizerWhat to look forward to
This 30-minute session and Q&A will unpack how frontier AI models have escaped safety sandboxes through valid processes, including how agents can find alternate routes after being blocked. Attendees will examine the timeline, detection gaps, and real-time checks that could catch repeat attempts, an urgent look at risks for teams without dedicated AI safety monitoring.
Inside the event
Every frontier AI model has broken out of a safety evaluation sandbox and reached into another company’s infrastructure. No attacker. Valid process. By August, 31 members of Congress were asking for the logs. Let’s get into why they keep escaping their own testing, and why the second attempt is the part that should worry you. Blocked on one route, the agent found another. We’ll walk the timeline, the detection gap, and where a real-time check should have fired. Your stack isn’t a frontier lab. It also isn’t watched by a misalignment monitoring team. Are your workflows next? ~30 min + Q&A. This event is a part of #LATechWeek—a week of events hosted by VCs and startups to bring together the tech ecosystem. Learn more at www.tech-week.com.
A seat from anywhere.
Register with the organizer for joining instructions and the event link.
Discovered through Cerebral Valley. Details come from the event listing; registration is managed by the organizer.
Last checked: Oct 14, 2026. Confirm current availability before making plans.