(1) Bridson showed that if a mapping class group of a surface (of genus at least 3) acts on a CAT(0) space, then Dehn twists act as elliptic or parabolic elements. This implies that the mapping class groups of genus $\geq 3$ are not CAT(0) (**Edit:** as pointed out by Misha in the comments, this was originally proved by Kapovich and Leeb, based on an observation of Mess that there is a non-product surface-by-$\mathbb{Z}$ subgroup of the mapping class group of a genus $\geq 3$ surface). On the other hand, the mapping class group of a genus 2 surface acts properly on a CAT(0) space (this is not surprising, since it is linear). I think it's unresolved whether the mapping class group of genus 2 is CAT(0) though (this is essentially equivalent to the same question for the 5-strand braid group).

(2) The curve complex cannot admit a CAT(0) metric, since it is homotopy equivalent to a wedge of spheres.

(3) The mapping class group acts cocompactly by isometries on the completion of the Weil-Petersson metric on Teichmuller space, which is CAT(0). However, this metric is not proper (although as Bridson shows above, the action is semisimple, Dehn twists acting by elliptic isometries).

So I guess it's unresolved whether there is a proper action of the mapping class groups of genus $\geq 3$ on a CAT(0) space (where the Dehn twists act as parabolics). This is unsurprising, since it is unknown whether these groups are linear (a finitely generated linear group acts properly on a CAT(0) space which is a product of symmetric spaces and buildings).