That was a bit tongue in cheek, it's probably the Foundation. The documentation doesn't specify really how it's done, only that the nodes "authenticate out of band" to determine state when necessary.
There's no key with special privileges. You're probably remembering "weak subjectivity," which is basically an acceptance of things like publicly known checkpoints every now and then. Anyone who's online continually doesn't need to rely on a checkpoint, but in PoS someone new coming in would need it...but this isn't all that different from needing to find out what software they need to run. E.g. Vitalik writes:
"It solves the long-range problems with proof of stake by relying on human-driven social information, but leaves to a consensus algorithm the role of increasing the speed of consensus from many weeks to twelve seconds and of allowing the use of highly complex rulesets and a large state. The role of human-driven consensus is relegated to maintaining consensus on block hashes over long periods of time, something which people are perfectly good at. A hypothetical oppressive government which is powerful enough to actually cause confusion over the true value of a block hash from one year ago would also be powerful enough to overpower any proof of work algorithm, or cause confusion about the rules of blockchain protocol."
How do you know which checkpoints to trust, unless they are signed?
And if you have trusted checkpoints I fail to see the point of proof-of-stake (or proof-of-anything really). Just call every block a checkpoint and you're done.
Any remaining use cases probably centers around availability, but we know how to build highly available systems and can build them to any degree. Trustless and decentralized systems are much harder.
