I don't think it's even that hard. Presumably an arbitrarily stronger intelligence could build arbitrarily subtle disaster-making flaws into whatever "helpful" technology/science it gives us. They could even have a generalized harmful sensation, as was discussed in another thread recently.