The Einstein field equations say ##G^{ab}\propto T^{ab}##. The left hand side describes curvature and its covariant derivative, ##\nabla_aG^{ab}##, is zero. This is called the Bianchi identity and is just geometry, like the sum of the angles of a triangle being 180° in Euclidean geometry. But it means that ##\nabla_aT^{ab}=0##, which is the law of local conservation of stress-energy that
@pervect mentioned. Roughly speaking, it says that the difference in stress energy in a region of space now and a tiny time later is equal to the amount that flowed out of the region in that time - or, stress-energy is neither created nor destroyed. All GR spacetimes respect this - they cannot avoid it.
The problem, also as pervect says, is that you can't generally take that local law and turn it into a global one (with honourable exceptions). I have the impression that opinion is divided over whether this is a problem and whether proposed solutions (if, indeed, it is a problem) are generally legitimate.