By their very nature Mashups are built on trust..

• You have to trust the API providers to keep that API up and running. Anyone that has done Twitter mashups know how much keeping an API up and running is worth.
• You have to trust the API providers to keep the API (somewhat) stable and to not remove features that you are depending on, or add features you really dont want. An example of this is all the mashups built on Google Maps, and we all know that there are way too many of those. Have the developers of those mashups really thought about what happens when Google puts ads on the Maps (and they will, they have already started in parts of the US)? I dont think so, the developers just trust Google to keep on going. If you have a commercial site using Google Maps you probably do not want ads for your competitors on those maps.
• If your mashup becomes a success you have to trust the API providers you are depending on not to change their terms of service so that your mashup suddenly becomes an illegal use of the API. Also, you have to trust the API provider not to use the fact that you are depending on their API for your business against you in a business negotiation.
• If you use web scraping you have to trust the site you scrape not to change to often (or at least trust your own ability to roll with the punches and update your scraping so it works with the new version of the site as well) etc. That is a lot of trust, and when mashups move from toys into real applications this becomes an issue.

All these things makes me a bit nervous, as I am a bit paranoid (note: this has not been clinically proven, I still think that THEY are out to get me). So how to mix a healthy bit of paranoia into my mashup building and get something good out of it all? What I do is that I always try to be aware of that I might have to switch API provider. Are you building a Twitter mashup? Why not also take a look at Jaiku’s API, or Pownce’s API (or Plurk or FriendFeed etc etc). You dont have to build your mashup so you can switch API provider in a matter of minutes, just be aware what else is out there so that you see the commonalities and don’t use to many features unique to one provider. This is the approach I am currently using when I am building mashups, at least I know that if shit hits the fan I can always go with somebody else. It will hurt a bit and take some effot, but I am not dead in the water. For the Google Maps example this would mean looking at Yahoo Maps and see what features the Yahoo Maps API have in common with the Google Maps API, and just use those common features. This can also come in handy if you hit the maximum number of requests on Google Maps, then it would be nice switch to Yahoo Maps automatically.

The risk with all this is of course to spend to much time preparing for something that won’t happen. It is the same situation as developers spending so much time making their code perfectly scalable and optimized that they acctually never ship anything in time. So dont go too far, but be aware of the situation. Trust is nice, but trusting several API providers to always do a good job and to not be evil in order for you to survive is quite risky.