My compact, one-liner solution…

print(np.loadtxt(sys.stdin, int).min(axis=1).max())

After I submitted this solution, it passed all the tests, and I began posting about it here, I realized something. I didn't skip the first line of input containing N and M as I usually do.

But my one-liner passed all the tests anyway! I'm just lucky that making those two values part of the array didn't make any difference in the results.

The solution I should've written was…

_ = input()  # We don't need N or M.  Skip them!
print(np.loadtxt(sys.stdin, int).min(axis=1).max())

Inspired by another user's comments about using N and M, although they're not necessary, I wrote a solution that uses them…

print(((NM := list(map(int, input().split()))),
    np.loadtxt(sys.stdin, int)[:NM[0], :NM[1]].min(axis= 1).max())[1])

Though, range M is not required, but its still always a good practice to do so to exactly respect the question asked. And that's why I intentionally invoked this to keep the N * M Matrix. What do you think? Please let me know your opinion?

import numpy

N, M = list(map(int, input().split()))

array_lst =[]

for _ in range(N):
    inp = input().split()
    lst = [int(inp[i]) for i in range(M)]
    array_lst.append(lst)
    
np_array = numpy.array(array_lst)
max_array= numpy.min(np_array, axis= 1)
print(numpy.max((max_array)))

Here is HackerRank Min and Max in Python solution - https://programmingoneonone.com/hackerrank-min-and-max-problem-solution-in-python.html

import numpy

n, m = map(int, input().split())

data = []

for _ in range(n): data.append(list(map(int, input().split())))

array = numpy.array(data)

print(numpy.max(numpy.min(array, axis=1)))