For some reason the article reads to me like “AI is not evil, it just has accidents when it loses coherence.” Sounds a lot like liability shifting.

They compared it to industrial accidents. I don't think a software company would try to shift liability by comparing themselves to factories explosions and chemical spills.