Can’t recall who said this, but the notion super intelligent AI will be inherently benevolent, as opposed to dangerous, is a fools bargain. At least that’s how I understand it.
He then mentioned that a super intelligence wouldn’t need much more than a browser to wreak havoc.