Note · Writing
Enterprises Need Their Own AI Kill Switch
The kill switch debate focuses on governments and frontier labs. If your organization lets AI act on its systems, you need a reliable way to stop it.
AI kill switches are getting a lot of attention, but the conversation is often focused on governments and frontier model providers.
I think enterprises need to think closer to home.
If your organization is giving AI access to systems, data, tools, or the ability to take action, you should also have a reliable way to stop it.
On a recent AI gateway build, we designed exactly that capability, including a simple rule: one authorized person could stop AI functions, but two had to agree to turn them back on.
Because “the AI did it” is not a valid excuse.
If you deploy it, you own the responsibility.