When I rent a safe deposit box from a bank, and store my valuables in it, my valuables do not become the property of the bank.
Likewise, just because I store some data on someone else's computer system does not make that data theirs.
Now, we may not yet have a good enough legal framework for protecting our data; in Europe, laws about this are better, while in the US you are expected to just use contract law for this purpose, and of course with online services you have no way of negotiating the contract and they always claim the right to do anything they want with the data.
But that does not make it "their data"; it just means that our laws need to be updated to better protect our data and not allow companies to simply claim they can do anything they want in non-negotiable user agreements.
You can argue that things should be different. What I am talking about is how things are.
The law around physical property in someone else's hands is on the other hand well established, and is referred by English speaking lawyers as "bailment". If you entrust people with your stuff and they misbehave, they are then liable for their negligence about it. So the legal theory should be simple, but "on the internet" is masquerading the issue. If I send some clothes to the dry cleaners to get cleaned, it is still my clothes. If I send some emails to be cleaned by a spam filter service, it suddenly is no longer my emails anymore. There is no reason why we need any new law to correct this, but there is political work to make it enforceable.
It's only when you add a shared construct like some system of ethics or law that your definition of data ownership make sense.
It's the difference between ownership defined by abilities and ownership defined by rights. It's a failure of the English language.
Computer security, especially, encourages you to think in base, practical (there must be a better word for this) terms. Since you can't guarantee other actors will respect your rights, you fall back to "If I don't want X done then I must make it physically impossible to do X rather than just wrong to do X".
Depends on the T&C [1]. Two things here, (1) you agreed to the FB terms and (2) took no steps to ensure your privacy despite these ever changing terms. More applicable digital lockbox hypo: when you put a gig of encrypted data on S3, AWS doesn't scan it and try to sell it to advertisers because you didn't agree to that. Further, they'd have to crack your encryption (i.e. PGP) as well if they wanted to violate those terms.
With AWS you get what you paid for. With FB, advertisers get what they paid for you.
(The set of things they can do with the data is subject to change, with notification, but if those changes are "adverse" to you, you have the right to revoke that consent. What happens to your data after that consent is revoked is something that probably needs to be explored legally, but it's pretty widely known that Facebook never actually deletes anything. And the TOS do say "irrevocable", after all...)
All that said, it's not their data. (Effectively) unlimited access, yes; ownership, no.
Actually, I disagree on that detail. An account on a Unix computer, accessible by SSH, is a known thing with known cultural implicit privacy guarantees. I have servers, which I can certainly call “mine”, with users, who have SSH access. I would absolutely not consider any file they type into to be “mine” — it is theirs.
Folks desperately want to deny the physical nature of cloud services, and pretend that its the same as storing something on your hard drive at home. They want to believe that the software abstraction that shows local and cloud documents together is real. But you can't will away reality.