If that's not libel, then the law has a bug, and needs to be fixed.
Regardless of the obscure legal technicalities, the facts are as follows: (1) Twitter is drawing a distinction between "safe" and "dangerous" sites; (2) Twitter is suggesting that the site in question is "dangerous," and (3) Twitter is causing the victim to suffer demonstrable losses as a direct result.
Twitter's act of drawing a distinction between "safe" and "dangerous" links should be what incurs the liability. They should not be able to wiggle out of it by using qualifiers like "potentially."
Everyone who has released software and had it marked by IE's SmartScreen filter with a lurid "Untrusted" warning dialog will understand this. Companies should not be allowed to do this without accountability.