As far as maintaining your data's integrity such that a retrieval program can have a shot at recovering it you have a few options. You can backup to a physical medium that is indestructible (within the timeframe and conditions you must decide yourself) or the medium can be meta-physically indestructible within the same conditions as above. This would appear to you as a black box since you do not need to concern yourself with the 'how' aspects of its implementation just the means of interfacing with it (physical and communication protocol) which would be done through any number of interfaces which as time progresses might change just as the definition and implementation of a "computer" may change.

Now we know that when we submit something to this black box it will be stored and as long as computers and computing don't change too drastically we be able to retrieve it at some future time. The black box portion of this could be in the form of a company that handles all of the details of looking to you like a black box as described for a reasonable fee but here you run into a problem. The company becomes a single point of failure that all of your data at risk. Perhaps governments could provide the black box duties but I, personally, have several problems with that option.

An alternative way of providing the black box functionality and eliminating the single-point-of-failure problem is to somehow, see that the open source community produces a multi-platform (easily migrated code is what we need here), distributed, well-designed and documented, open source, peer-to-peer, backup program. I know that the topic has been discussed here before but don't know the current status of any of the projects.

This would be a great project for the open source community as it would solve a problem that has been faced by anyone that deals with large amounts of data and doesn't have a fortune to spend keeping it retrievable let alone easily accessed and navigated. There are many products, both software and hardware, that provide or, at least, approach meeting the requirements of the overall solution.