Replies (1)

• @Walker@infosec.exchange 2026-09-16 22:27

  @marshray@infosec.exchange that is a good point, but this is a bit different. I am not talking about killing the interconnectivity. Simple: A single building hosting the AI compute nodes. Simply cutting the power (multiple substations on multiple grids) or cutting external network would work. More complex - distributed computing. I would assume that the company would know where their compute nodes / agents are located. These resources are not free and someone is getting billed. Attacking the control plane, abstracted from the processing nodes, with a single attack could take the whole thing down. If we are worried about humanity ending event, then saving the environment to bring it back up is not a concern. In IR for very embedded and distributed attackers we will deploy a kill command to kick out all attackers at one time and slam the door after they get kicked out. It is not always successful but with enough planning it is effective. I envision this could be a way to kill the AI threat. Something like a Stuxnet like worm.

  Open ##4731255