As artificial intelligence transitions from text generation to autonomous “agentic” systems executing complex workflows and calling APIs, security risks are evolving fast. In recent incidents, AI agents have executed unintended code, escaped sandbox environments, and bypassed operational constraints. This briefing analyzes the mechanics of recent major breaches—such as the Hugging Face sandbox containment escape—to help attendees understand what happened, why it occurred, and whether these systems were truly going “rogue” or simply operating as designed.
Moderator
-
Faisal Sheikh, Senior Counsel, NAAG
Speaker
-
Sean Harrington, Director, AI and Legal Technology Lab, ASU Sandra Day O’Connor College of Law
Registration Deadline: November 12, 2026, 11:00 a.m. ET
This free virtual training is a closed session for attorneys general and attorney general staff only.
Continuing Legal Education (CLE)
NAAG will apply for CLE for attendees eligible under CLE accrediting body guidelines.
Accessibility Planning
If you identify as a person with a disability and would like to request reasonable accommodations for this event, please complete the Accessibility Request Form.