I suspect these controllers were mostly being used for telemetry (getting notified if they lose power), not remote control. But they were still vulnerable to zero-days if not updated. Kind of the same story as all the other IoT stuff that has turned out to be vulnerable.

I think LLMs are fine for specific uses. A useful technology for brainstorming, debugging code, generic code examples, etc. People are just weary of oligarchs mandating how we use technology. We want to be customers but they want to instead shape how we work, as if we are livestock