/ About this 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.
