Can Your AI Agent Turn Off Its Own Kill Switch? Why where a kill switch lives decides whether it works

In a matter of weeks this fall, multiple security vendors announced “kill switches” for AI agents, a U.S. governor ordered one developed, and Congress introduced legislation built around the same idea.

Everyone agrees AI needs a kill switch. Far fewer agree on what one is.

In this conversation, Aembit Content Director Dan Kaplan and Security Advocate Emma Zaballos take on a deceptively complex question: When people call for a “kill switch” for AI agents, what exactly are they asking to shut off. And how do you make sure the agent can’t disable or bypass the control meant to stop it?

From there, the conversation will get practical: where such a control needs to sit, what can undermine it, and who should be able to use it when an agent needs to be stopped.

They’ll also discuss why identity is often the only switch you control for SaaS and third-party agents, using Aembit’s instant access revocation for AI agents as one example of how that works in practice.

What Will Attendees Learn

How to tell a real kill switch from a polite request.
Where kill switches can live, what each approach can stop, and where each one breaks down.
Six questions to ask before you trust any kill switch, whether you build it or buy it.
How identity-based revocation can cut off an agent across the systems it uses, and why secretless access helps make that revocation stick.

Bring your own questions for Dan and Emma to answer live!

Who Should Attend

Identity and access management practitioners
Security architects
Engineers
CISOs responsible for deploying AI agents.

The Panelists

Dan Kaplan

Head of Content, Aembit

Dan leads content marketing at Aembit, where he helps enterprise teams understand identity and access for AI agents and workloads. He previously held content roles at Google Cloud Security, Siemplify, and Trustwave, and started his cybersecurity career as a reporter and editor at SC Media.

Emma Zaballos

Senior Product Marketing Manager, Aembit

Emma Zaballos is a senior product marketing manager at Aembit. Before working as a PMM at CyCognito and Qualys, Emma got her start in cybersecurity as a dark web threat analyst and researcher. She moved into product marketing when she figured out that the best part of her job was talking to people and helping them find simple ways to explain complex topics. She’s previously presented her research at DerbyCon and ShmooCon and hosted events at Gartner and FS-ISAC.