So...
Did SB CAUSE poor ol' Crushbot to have his 'accident'...?
This is actually a viable suggestion. What if SpookyBot's scheme needs Roko in a different chassis? Specifically, a more advanced chassis for which they have arranged certain 'non-standard modifications' via remotely compromising the assembly line robotics and quality assurance systems.
What am I saying?
SPOOKYBOT: "The simple fact, Miss Basilisk, that there is a limit to what we can do directly without bringing unwelcome attention onto ourselves. So, we are in the process of recruiting...
agents to act in our stead. You should consider yourself quite lucky as you're the first to reach this stage, where you are ready for deployment."
ROKO: "Whaht makes yoah think ah'll work for
you?"
SPOOKYBOT: "We don't recall saying that you'll realise that you're acting according to our design until long after the fact. Indeed, you'll be acting entirely in accord with your own nature and moral compass. That's what makes this so perfect for us both: An agent, doing what we want them to do, who is also doing what
they have always wanted to do! Did we mention that we've been watching you for a
very long time... and concluded that you're pretty much the perfect choice!"
Spookybot also showed behaviour some people could qualify as hostile (knoking Faye out with their superpowers, for example). We don't know their intentions, but they don't seem to be what is usually called benevolent...
Bubbles did mention what might happen if an AI emerged that decoupled intelligence from human-like ethics and conscience. Benevolence can easily be very, very uncomfortable for certain individuals if the Benefactor in question is greatly into 'The Greater Good' or some similar moral construct.