Tahoe-LAFS providers like Least Authority [1] and Matador Cloud [2] pride themselves on not being able to access your data.
With Tahoe LAFS, you're either downloading the pre-built binary or compiling from source. Either way, you're trusting the person who signed the binary/source or the person who hosts them.
This is similar to when Apple says they can't read your messages. Sure, they may not able to decrypt the data on their server, but they're in total control of the client software. They can get access to your messages.
I do think it's in Tahoe/Apple/etc's interest to NOT be able to see their data. For one, they said they can't and reputations are important. It may also be beneficial when dealing with law enforcement requests. So trusting them isn't totally unwarranted, but there's a lot there that isn't mathematics.
(You're never going to get to 100% math, but there are ways to get further. For example, someone might eventually write a Tahoe LAFS client that comes with a machine-checkable proof that it doesn't leak plaintext. You now need to trust the proof checker, but it's progress.)
I believe you can architect a solution such that it's easy enough to get a third party (or yourself) to verify it is safe and private, even in the face of a hostile backup provider (or more realistically, a backup provider that's met an opportunistic law enforcement agency waving a sternly worded letter).
If enough parties collude, they can still gain access to your data, but at that point I'm pretty sure the backups won't be the weakest link anymore.
I mean, you're still running an OS, and compiling with a compiler, installing and running software from a package distributor, and using a CPU with a management engine, and all those things might have backdoors too.
For example, people should know that blanket statements like "Apple cannot read your messages" are false. And your client/server protocol can be state of the art, but you still rely on the much-maligned HTTPS certificate infrastructure (among other things) to get the client bits.
If this space is interesting to you, spending a few days reading through their Trac is educational and rewarding! Highly recommended.
Trivia: One of the founders of LAFS went on to found Zcash, the (actually) zero knowledge crypto currency.
I'm a paying customer and I use SpiderOak for some non-critical backup/synchronization tasks but I refrain from using it for anything sensitive because of this.
I'd also like to be able to hack the client to remove all the cruft I don't need, it's not exactly pleasant to use currently. On linux I'd actually prefer a good CLI over the weird GUI we have now.
FYI, many people do use SpiderOak exclusively from the command line. It's fairly scriptable.
Here's the command line options: https://spideroak.com/faq/how-can-i-use-spideroak-from-the-c...
Also, the source code for our other products (Encryptr, Semaphor) is published. SpiderOakONE was first developed in 2006 and open source business models were not so popular at the time. It's been much harder than I thought it would be to make SpiderOakONE open source but one glorious day we will get there. https://spideroak.com/solutions/semaphor/source
If their privacy policy prevents them from accessing your data unless authorized, then this isn't an unreasonable statement.
Often a company needs data for a certain purpose, but there is no good technical solution for enforcing that. This is where privacy policies can benefit users, by providing a legal agreement to protect users beyond the abilities of the underlying technology.
Your typical user is much more at risk with end-to-end encryption than with a well-architected SaaS product, since their likelihood of being a victim of some sort of easily preventable fraud is much more likely than them having their data misappropriated. (Think of how Gmail automatically handles con artists, phishing scams, viruses, XSS attacks, SPF/DKIM/DMARC, etc.)
Because of this, users tend to be better off having their data protected through a combination of technology and legal agreements rather than protected solely through technology, although this can be difficult to communicate.
E.g. how would you communicate to your (archetypal) grandmother that she's better off using Gmail than the same sort of setup as Edward Snowden, and make her feel safe doing so?
Not all e2ee products are created equal. I think your description is accurate for the PGP ecosystem, because for example it's hard to be sure you've got the right key from the key servers, and anyone who can use email can contact you. In fact most crypto folks I know believe email is an unsecurable platform.
However good e2ee systems provide strong Authentication of the data's origin, which is often more valuable than the encryption. In SpiderOak's Semaphor for example, a team admin approves new team members before they can communicate with other people on the team. It's explicitly a tool intended for safe internal communication with a specific in group, such as an enterprise team.
If I'm running your software, and there exists any channel by which that software can talk to you, then ultimately you can access whatever that software has access to.
This is true about any entity that has access to your data, including your bank, health insurance, or any other organisation which you have an account with. It is true that they don't say they can't access your data though...
Regarding their motivation - having developed a system with end to end encryption for sensitive information, I can tell you that the worst thing from both a business and professional perspective would be a data security breach - that gives quite a lot of motivation to get things right...
Am I running your software in a process that has network access? Then you can access my data.
I understand the point you're trying to make, and I totally get that architecting a system so that unencrypted data doesn't leave my device is superior to an architecture where it does.
But I still must, ultimately, trust you, your competence, and your motivations. If I trust that you don't want to access my data, and have tried to architect your systems so that is hard to do, and are competent to do so, then I can trust my data is probably safe.
But that's not the same as it being physically impossible for you to access my data.