The sum of the kinetic and potential energy of a bound system is negative. This must be the case, because you would have to inject more energy into the system to separate the components to infinity.
Therefore the "gravitational mass" of the binary - what you would measure with another orbiting test mass at greater distance - would be less than you would expect from the sum of all the masses of the components of the system measured when they are far apart and stationary.
Whether this is an important effect (i.e. the relative size of the correction) can be judged from the ratio of the absolute value of binding energy to the rest mass energy:
$$\alpha \simeq \frac{GM_1M_2}{2Rc^2(M_1+M_2)}\ ,$$
where $R$ is the separation and the result is exact for a circular orbit.
The effect can be important in binaries featuring compact stars (e.g. neutron stars) that have stellar masses and where $R$ can be quite small - the ratio above is a few per cent for a pair of neutron stars separated by 30 km.
As an aside, this consideration also applies to single stars, where the sum of their internal kinetic energy and gravitational potential energy is also negative. Again, this is important in white dwarf and neutron star physics.