I've argued this before around here - I believe it's a platform provider's responsibility for the most part to secure data, and less so the responsibility of the developer or user. Amazon should go much further and make it hard to open-up the data to the public, at least for certain categories of buckets.
So for instance some buckets should always be public by default, and some should always be encrypted and private by default. That should make intelligence agencies' choice easier, because I would imagine even if it's "harder" to process the data from an encrypted bucket, they would still prefer that option to the always public bucket.
And maybe both categories could still be configured to either be private or public, respectively, but the account owners should have to really go out of their way to make those changes. So most shouldn't bother, and just use the defaults for each category of buckets.
I would say that is going too far, or maybe I'd say it differently. If a certain problem becomes very frequent relative to it's severity, the problem is a design bug and not just user error. The provider isn't responsible for every mistake, but they are responsible for designing with mistakes in mind.
I think we are way far away from users fully understanding the risks, and we're still mostly dealing with people not realizing they're vulnerable. So I put this primarily on the provider.
I wonder what other improvements we'll see when CIA's surveillance or drone strike data is also exposed to the public by a similar fuckup? Fully homomorphic encryption?