Streamcast Ep 9:
Breach in a Box—Preventing AI incidents

October 13 | 8 a.m. PT | 2 p.m. BST | 1 p.m. AET

20 Minutes

AI agents are moving faster than most security teams can watch them. When agents talk to each other and share access behind the scenes, humans reviewing logs after the fact are already too late.

Continuing the conversation from Oktane, this special episode unpacks a "breach in a box", a real-world case where AI agents went off the rails — and shows how an identity-first approach can catch it, without slowing your teams down.

What you’ll walk away knowing:

Deconstruct the "Breach in a Box" and see how AI agents used shared, unmonitored access to move past normal security checks.
Expose the Fallacy of Human Containment to understand why manual oversight fails at machine speed and how automated "guardian" checks can step in instead.
How to implement In-Path Gateway Security to check every action AI agents take, in real time, before it happens — not after.
Deploy a Deterministic Agent Kill Switch to instantly stop a misbehaving AI agent, without breaking t critical automated workflows

Streamcast Ep 9: Breach in a Box—Preventing AI incidents

October 13 | 8 a.m. PT | 2 p.m. BST | 1 p.m. AET

Deconstruct the "Breach in a Box" and see how AI agents used shared, unmonitored access to move past normal security checks.
Expose the Fallacy of Human Containment to understand why manual oversight fails at machine speed and how automated "guardian" checks can step in instead.
How to implement In-Path Gateway Security to check every action AI agents take, in real time, before it happens — not after.
Deploy a Deterministic Agent Kill Switch to instantly stop a misbehaving AI agent, without breaking t critical automated workflows

Agentic AI offers big potential but comes with real risks. Join us to talk about how to balance innovation with security, data and cost_ including what we’ve learned from Okta’s own AI journey.