A session about boundaries: what happens when an agent has too much autonomy, too much tool access, or too little supervision. Framed around the idea that good agent design is really about deciding what an agent is NOT allowed to do. Strong fit for a security- or governance-adjacent track.

AUDIENCE
Platform teams, security engineers, and architects responsible for agent governance, approvals, and access control — not just the builders.
EXPECTED OUTCOMES
• Leave with a concrete checklist for scoping agent permissions before launch
• Understand the human-in-the-loop pattern and when it's mandatory vs. optional
• See a live example of an agent escalating correctly vs. incorrectly