How exactly (and why) can I access all of np through numpy.random?

I was poking through what is available in numpy.random after importing

from numpy import random 

with dir(random), and noticed that there was a variable np in scope, which appears equivalent to the top-level module, for example

In[1]: from numpy import random

In[2]: random.np.fft.fft2
Out[2]: <function numpy.fft.fftpack.fft2>

In[3]: random.np.random.np.fft.np # not the same for fft

AttributeErrorTraceback (most recent call last)
<ipython-input-78-a64e04c36c80> in <module>()
----> 1 random.np.random.np.fft.np

AttributeError: module 'numpy.fft' has no attribute 'np'

This seems a bit strange to me... or at least not something I've seen before in other Python modules. It looks like I can access everything I could with import numpy as np through the np variable in random.

I wanted to see how it was available to the submodule, so I looked in numpy/random/__init__.py in the source code, and didn't see how it is made available. I also looked in numpy/random/info.py at __all__, but can't find how it is exposed to the module.

How is the top level module made available to numpy.random, and is there any motivation for having it available?