We say and write what we observe. In history, when an axisymmetric cross-sectional beam (i.e. the beam's cross-section still remains the same when rotated about its longitudinal centroidal axis) was subjected to torsion, the cross-sections of the beam remained plane/flat. This is true for both, solid beam and hollow beam. However, when a non-axisymmetric cross-sectional beam (i.e. the beam's cross-section DOESN'T remain the same when rotated about its longitudinal centroidal axis) was subjected to torsion, it was observed that the cross-sections didn't remain plane/flat. In fact, they warped when subjected to twisting. The general equation for calculating stresses due to pure torsion for a solid circular beam is shown below:
![enter image description here](https://cdn.statically.io/img/i.sstatic.net/Xr9wN.png)
where T: Torque applied, r: distance from the centroid, J: Polar moment of inertia. Now, the derivation of this equation involves an assumption that the plane section must remain plane, i.e. the cross-section cannot warp. Therefore, this equation cannot be applied to any other non-axisymmetric cross-sectional beams.
Warping basically refers to one-half of the cross section (above/below the neutral axis) being subjected to compression, and the other half to tension. This behavior is what we usually see in a beam subjected to bending, however, this behavior is also observed in non-axisymmetric cross-sections subjected to torsion only. Calculating shear stresses for a non-circular cross-section is somewhat more complex and complicated than the circular ones.