The title, along with the rest of the article, seems to obscure an important point. Although the entangled particles appear to communicate with each other FTL, the observers can't communicate any information with each other this way. There are still lots of applications of this to quantum computing, but there wouldn't be any way to build an ansible with this.
The prohibition of faster than light travel (or communication) in GR only applies locally. That is to say, it really should be phrased as "two observers can not have a relative speed faster than c at the same point."
If you're not talking about the same point of space, the space between the observers could be expanding and the rate of this expansion can legitimately be faster than c without violating the relativity principle.
The trick to creating genuine FTL travel or communication is to somehow multiple paths between the same points in spacetime -- i.e. a wormhole. If the new path is shorter than the other, then information can travel along it at a speed less than c, but still take less time than light going the long way.
Quantum entanglement whole different game. Once a measurement has been made, the entanglement is broken and the components are left in a particular state. But, while the observer that did the measurement knows the state of the other component, they can only convey that information via regular, light speed, communication.
"If the new path is shorter than the other, then information can travel along it at a speed less than c, but still take less time than light going the long way."
If light can go through that wormhole, too, it technically wouldn't be FTL at all, would it? Otherwise, we already have FTL travel: point a laser at a mirror on the moon, step a meter aside, and wait for the light to finally get there.
I'm not certain, but the abstract says We prepare the teleporter through photon-mediated heralded entanglement between two distant electron spins. So I think there's a classical setup phase, then the information transfer is instantaneous.
EDIT: After reading the article, the novelty there is that they entangled particles separated by 3 meters. They didn't achieve any unusaly long entanglement time. That won't (or, at least, it's not trivial how it would) lead to better quantum computers, but could lead to qbit transmission and cloud quantum computers, as the authors say.
The original:
From the abstract, they found a way to "freeze" entangled qbits, so that you can transport them, and read later.
For the confusion here, just remember that you can not transmit information by reading quantum entangled pairs. You can use them to make a quantum cryptography key exchange, that's still as subject to man in the middle attacks as symetric crypto, and you can use them for error correction in quantum computers (for what this one may be HUGE).
