Risk from competent malign agents - AI agents could be competent and want different things from humans, which could be bad for us

Expert opinion - Many AI experts are worried about the destructive capabilities of AI

Catastrophic tools - AI might lower the cost or knowledge required to make catastrophic weapons

Loss of control via speed - Additional speed makes many situations more dangerous. AI is developing very fast

Large impacts suggest large risks - AI's impacts are likely to be very large

Black boxes - We don't understand how AI systems work, so they might be very destructive

Second species argument - Apes should have been wary of "inventing" a species more intelligent than themselves, so should we

AI may produce or accelerate destructive multi-agent dynamics - There may not be a stable equilibrium with AI in it

Loss of control via inferiority - A child heir to a large fortune could struggle to know which adults to trust and so slowly lose control of it

Humans aren't aligned to one another - Given full control, most people's utopias are incompatible. AI isn't new here, but it gives more power to achieve "utopias" that others might find horrific