This may have changed, but the last I heard was that Topix relied heavily on automated filtering and attempted to remove only the most abusive posts that got past the filtering, e.g., death threats. Their philosophy was, more or less, "If someone calls you a pathetic moron and likely spouse abuser, we won't remove the post but we will give you free rein to defend yourself and insult your attackers in return."
From a community building standpoint, I think that sort of approach encourages flame wars and discourages participation by thoughtful individuals.
Trying to develop a scalable moderation model for their number of location sites (2000+) and six-figure daily posting volume is a real challenge.
moTi, websites which allow user contributions in the US have generally been afforded legal protection as service providers. The author of libelous content (i.e., the poster) is still liable, and the site owner can be forced to divulge registration information, IP addresses, and so on, by a court order.
Still, the vast majority of site owners try and remove libelous content quickly, both for community reasons and to avoid getting tangled up in messy and expensive lawsuits.