Can we determine whether or not a particle is entangled? Suppose Shaniqua and Tyrone have four pairs, a, b, c, and d, of entangled particles. They take their particles and go very far apart. If Tyrone can determine whether or not a particle is still entangled, Shaniqua could observe, for example, a and c, transmitting the binary number 1010 faster than light. So, can we determine whether or not a particle is entangled? If so, why doesn't it lead to faster-than-light information transfer?
 A: Entanglement is just a correlation in measured properties of the subsystems (particles) expressed in a quantum way. You may only determine whether properties are correlated (or entangled) in a given, "initial" state if you repeat some measurements of the system with the same initial state many times. If you only measure two spins, for example, once, you get some result, like up-up or up-down or down-up or down-down but none of the four possibilities is more or less entangled than others. All of them may occur in entangled initial states and all of them may occur in non-entangled initial states.
Entanglement only means "predicted properties of the two subsystems are correlated, moreover correlated in a way that isn't captured by a simple classical model of correlation". Whether the predicted properties are correlated may be determined from the probability distribution(s) but to measure the probability distribution(s), you have to repeat the experiment with the same initial state many times. 
More precisely, entangled states are those that can't be written as a tensor product of wave functions describing the separate subsystems. Once at least one of the entangled variables is measured, the entanglement becomes meaningless because the value of the variable is suddenly known and we're only left with some general wave function for the other, previously entangled variable which remains unknown up to the second measurement (this reduction of the dependence of the wave function is misleadingly referred to as the infamous "collapse"). And if there is only one variable, it can't be entangled.
But nothing physical is changing about the variable that hasn't been measured yet. The overall probability distribution for various outcomes $y$ remains the same after the first measurement of the (faraway) variable $x$ is performed (imagine it's a probabilistic distribution $\rho(y)=C\rho(x_\text{just measured},y)$ that is left afterwords, $C$ is such that $\int dy\,\rho(y)=1$). So no information can be transmitted by the fact that the first measurement took place.
Quite generally, quantum mechanics doesn't need any genuine (one that could transfer useful information) faster-than-light communication to guarantee things such as correlations of measurements done with entangled states. And in relativistic, local theories – especially quantum field theories and string theory – one may prove completely generally that a superluminal transfer of information is not only unnecessary for quantum mechanics to work; it is actually prohibited and impossible, by the basic laws of special relativity.
Addition 2017:
One cannot measure "whether two subsystems are entangled" by a single measurement, I wrote. Mathematically within quantum mechanics, this is equivalent to saying that there is no observable that would act as $0$ on non-entangled states and $1$ on entangled states. It's simply to see. The first condition says that the operator $E$ (entangled or not) annihilates all factorized basis vectors $u_i \otimes v_j$ because those are not entangled. But observables in quantum mechanics must be given by linear operator which means that $E$ annihilates all the linear combinations of states $u_i \otimes v_j$ as well – it acts as $0$ on the whole Hilbert space. So it just cannot act as $1$ on entangled states. End of proof.
Within the ER=EPR correspondence, this simple conclusion means that one can't also completely reliably measure whether a traversable wormhole is present in a given state. Especially when such would-be Einstein-Rosen bridge were very small, it would run exactly to the problem above. One may at most measure whether the state of the system is compatible with a particular entangled state. But within ER=EPR, all states may be built by either taking the tensor products i.e. non-entangled states without wormholes, or as modifications of a particular entangled state – a wormhole. These are two allowed descriptions of the same "composite" system or a particular wormhole and we can't say and we can't measure which one is right. Both of them are always equally right. It's really just another application of Bohr's complementarity. A system, in this case composite system, is ready to be observed in many ways – many observables may be measured – that can't be measured at the same moment but all these incompatible schemes may be extended to a complete set of observables etc.
A: What if you have a grouping of multiple entangled particles e.g. 4 photons entangled as one. you split up the particles and spread them over space. on observation all the particles should have the same spin and after than point they should become untangled. So an observer can observe if a flag has been set at some point in the future by checking particle state. 
If you have a very large number of these particles what would stop of you from periodically checking sentinel particles for an untangled state and as soon as this is witnessed engaging in bidirectional communication by selectively detangling groups of particles waiting for the next sentinel and then reading another group of particles from the origin source?
Going back to the risk of particles being randomly untangled what would prevent some application of statistical analysis and additional particles being added to the mix to allow better signal/noise quality. 
